Intel SL3QA - Pentium III 550 MHz Processor, SL6BY - Pentium III Processor 1.4GHz Tualatin 512KB/133/370, Pentium III Specification

Download

Intel® Pentium® III Processor

Specification Update

August 2008

Revision 060

Document Number: 244453-060

INFORMATION IN THIS DOCUMENT IS PROVIDED IN CONNECTION WITH INTEL® PRODUCTS. NO LICENSE, EXPRESS OR IMPLIED, BY ESTOPPEL OR OTHERWISE, TO ANY INTELLECTUAL PROPERTY RIGHTS IS GRANTED BY THIS DOCUMENT. EXCEPT AS PROVIDED IN INTEL’S TERMS AND CONDITIONS OF SALE FOR SUCH PRODUCTS, INTEL ASSUMES NO LIABILITY WHATSOEVER, AND INTEL DISCLAIMS ANY EXPRESS OR IMPLIED WARRANTY, RELATING TO SALE AND/OR USE OF INTEL PRODUCTS INCLUDING LIABILITY OR WARRANTIES RELATING TO FITNESS FOR A PARTICULAR PURPOSE, MERCHANTABILITY, OR INFRINGEMENT OF ANY PATENT, COPYRIGHT OR OTHER INTELLECTUAL PROPERTY RIGHT. Intel products are not intended for use in medical, life saving, or life sustaining applications.

Intel may make changes to specifications and product descri ptions at any time, without notice.

Designers must not rely on the absence or characteristics of any features or instructions marked “reserved” or “undefined.” Intel reserves these for future definition and shall have no responsibility whatsoever for conflicts or incompatibilities arising from future changes to them.

The Intel® Pentium® III processor may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request.

Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order.

Intel, Pentium, Intel Xeon, Celeron, MMX and the Intel logo are trademarks or registered trademarks of Intel Corporation or its subsidi aries in the United States and other countries.

*Other names and brands may be claimed as the property of others.

2 Specification Update

Contents

Preface.............................................................................................................................8

Summary Tables of Changes.............................................................................................. 10

Identification Information.................................................................................................. 20

Errata ............................................................................................................................ 39

Specification Changes....................................................................................................... 88

Specification Clarifications ................................................................................................. 89

Documentation Changes ................................................................................................... 93

Specification Update 3

Revision History

Revision Description Date

-001 This is the first Specification Update for Pentium® III processors. March 1999

-002 Added Erratum E42. Deleted Erratum E16 and renumbered existing items. Corrected Errata table “Plans” column for E39. Updated the Pentium III Processor Identification Information tab le.

-003 Updated the Pentium III Processor Identification Information table. Updated the Errata table by marking Errata E34, E35, and E40 as Fixed.

-004 Updated the Pentium III Processor Identification and Pack age Information table. Added Erratum 43. Added Documentation Change E1. Added Specification Clarifications E1 and E2. Added Specification Change E3.

-005 Added footnote 4 to the Pentium III Processor Identification and Package Information table. Added Erratum E44. Added stepping Kc0 in Summary Table of Changes. Added Mixed Steppings in DP Systems section. Updated Documentation Changes, Specification Clarifications, and Specification Changes introduction paragraphs.

-006 Added Errata E45 and E46. Added Documentation Change E2. Updated Identification Information table. Updated and corrected Pentium III Processor Identification and Package Information table. Updated Codes Used in Summary Table. Updated column heading in Errata, Documentation Changes, Specification Clarifications and Specification Changes tables.

-007 Revised Errata E45. Updated DP Platform Population Matrix for the Pentium III Processor with 100 MHz System Bus. Updated datasheet references to include the latest supported frequency.

-008 Added Errata E47. Updated the Pentium III Processor Identification and Package Information table. Added the DP Platform Population Matrix for the Pentium III Processor with 133 MHz System Bus table. Added Brand ID column to Identification Information. Updated datasheet references to include the latest supported frequency.

-009 Added Errata E48 and E49. Added Documentation Change E3. Added new stepping column in the Summary of Changes tables. Updated the Pentium® III Processor Identification Information tables. Up dated Mixed Step pings in DP System section. Updated the Pentium® III Process Identification Information table. Updated references.

-010 Updated document references in Preface to include new Pentium III processor datasheets. Updated errata E10, E11, E19, and E32 in the Summary of Errata table. Added Errata E50-E58. Added Documentation Change E4. Added Specification Clarification E3. Added Specification Changes E4 and E5.

April 1999

May 1999

June 1999

July 1999

August 1999

September 1999

October 1999

November 1999

December 1999

4 Specification Update

Revision Description Date

-011 Corrected an error in the Summary of Errata table. Erratum E56 was incorrectly shown as applying to the Ca2 stepping. Erratum E56 does NOT apply to the Ca2 stepping.

-012 Updated Preface to include new Pentium III processor datasheets. Added 800-MHz Pentium III processor information to the DP Platform Population Matrix tables and the Pentium® III Processor Identification and Packaging Information table. A dded note 10 to the Pentium® III Processor Identification and Packaging Information table and updated Notes column and other table data. Updated erratum E51. Added Errata E59-E62. Added Documentation Change E5. Added Specification Change E6.

-013 Updated Errata E49 and E61. Added Documentation Change E6. Updated the Pentium® III Processor Identification Information. Updated S-Spec SL365. Updated Summary of Changes product letter codes.

-014 Updated Preface to include new Pentium III processor datasheet. Updated Pentium® III Processor Identification and Package Information table. Updated Summary of Errata, Summary of Documentation Changes, Summary of Specification Clarifications Summary of Specification Changes tables with Cb0 stepping. Updated Erratum E48.

-015 Special Launch Edition: Updated the new Cb0 stepping information. Updated the document references in the Preface. Updated DP population table.

-016 Updated Processor Identification Information table. Updated DP Population Tables. Added Errata E63 & E64.

-017 Updated Pentium III Processor Identification and Package Information table. Updated Errata E64. Added Errata E65 & E66.

-018 Updated Processor Identification, Summary of Errata, and Summary of Specification Changes tables. Updated Dual Processor Tables. Added new Specification Change E7.

-019 Added new errata E67 & E68. Updated Processor Identification Table.

Edited erratum E36. Updated Processor Identification, Summary of Errata, Summary of Documentation Changes, Summary of Specification Clarifications, Summary of Specification Changes tables with cC0 Stepping.

-020 Added new Erratum E69. Updated Dual Processor Matrix. Updated Dual Processor Matrix. Updated Processor Identification Table with new C0 step CPUs.

-021 Added New Errata E70 & E71. Added Re-Writes for Errata E28, E48, & E62. Added New Documentation Changes E7 & E8. Updated Dual Processor Matrix, removed TBDs. Updated Processor Identification Table.

-022 Updated Pentium® III Processor Identification Information table. Updated Dual Processor Matrix. Added New Errata E72 and E73. Added New Documentation Changes E9 and E10.

-023 Updated Processor Identification Table. Added New Erratum E73. November 2000

December 1999

January 2000

February 2000

March 2000

April 2000

May 2000

June 2000

July 2000

August 2000

September 2000

October 2000

Specification Update 5

Revision Description Date

-024 Updated Specification Update product key to include the Intel® Pentium® 4 processor, Revised Erratum E2. Added new Documentation Changes E11 – E16.

-025 Revised Erratum E2. Added new Documentation Changes E17 and E18. Updated Processor Identification Table.

-026 Added new Documentation Change E19. Revised Documentation Change E17.

-027 Added new Errata E74 and E75. March 2001

-028 Added erratum E76 March 2001

-029 Revised Erratum E76 to Fixed. Updated processor identification table. Updated the tables in the Mixed Steppings in DP Systems section.

-030 Updated note 18 in the Pentium® III Processor Identification and Package Information table. Updated Specification Update product key to include the Intel® Xeon™ processor

-031 Special Launch Edition: Added package marking information under General Section. Added new S spec info into processor table. Updated Dual Processor tables. Updated Summary of Errata, Summary of Documentation, Summary of Specification Clarifications, and Summary of Changes tables. Added Errata E77 & E78.

-032 Added new S spec info into processor table. Revised package marking information under General Section. Updated Dual Processor tables. Deleted duplicate information in the processor ID table.

-033 Added new errata E79 and E80. Updated DP Matrix Tables August 2001

-034 Changed word “motherboard” to “baseboard” in erratum E78 October 2001

-035 Updated DP population matrix for new tA1 parts. Added Doc Changes E20, E21, E22, E23, and E24.

-036 Special launch edition. Added part with S-Spec SL5VX at 1.33GHz to the Pentium® III Processor Identification and Package Information list.

-037 Added: Doc changes E25, E26, E27, E28, E29, Spec Clarification E4, E5 and Spec Change E8. Added part with S-Spec SL657 to the Processor ID Information Table

-038 Modified Erratum E80 and added Erratum E81. Added Doc Change E1. March 2002

-039 Out of Cycle Special Launch Edition. Added Server LV part with S-Spec Number SL66D at 800 MHz to the Pentium® III Processor Identification and Package Information list.

-040 Added Doc Change E1. April 2002

-041 Modified Errata E59. Added Doc Changes E1, E2, and E3. May 2002

-042 Added Erratum E82. Added Doc Changes E1 and E2. Added parts with S-Spec numbers SL6C2, SL6C3, SL6C4 and SL6BZ with Core Stepping tB1 to the Processor ID Information Table.

-043 Added Document Changes E3-E12. Added parts with S-Spec numbers SL6BW, SL6BX, SL6BY and SL6HC with Core Stepping tB1 to the Processor ID Information Table. Changed status of erratum E78 to NOFIX.

December 2000

January 2001

February 2001

May 2001

June 2001

July 2001

November 2001

February 2002

March 2002

June 2002

July 2002

6 Specification Update

Revision Description Date

-044 Added new Doc Changes E3-E24. Removed parts with S-Spec numbers SL6C2, SL6C3, SL6C4 and SL6BZ with Core Stepping tB1 from the Processor ID Information Table.

-045 Added Doc Changes E25 to E32. Updated Summary of Changes. October 2002

-046 Updated DP FC-PGA2 Matrix table. Added one S-spec number. December 2002

-047 Added Erratum E83. March 2003

-048 Updated Summary of Errata table E78 and Added one S Spec number. May 2003

-049 Added Errata E84 and E85. November 2003

-050 Updated E85 and added E86 to E89 October 2004

-051 Added Errata E90, E91, E92 November 2004

-052 Added Errata E93, E94 February 2005

-053 Updated Errata E80 March 2005

-054 Added Specification Clarification April 2005

-055 Added Errata E95, E96, E97, E98, E99, E100, E101, E102, E103, E104, E105, E106, E107. Updated Summary Table of Changes. Updated the name of the Sofftware Developer Manuals.

-056 Added Erratum E108 January 2007

-057 Updated Summary Table of Changes. May 2007

-058 Added Erratum E109. Updated Summary Table of Changes. August 2007

-059 Updated Summary Table of Changes. April 2008

-060 Updated Summary Table of Changes. August 2008

September 2002

December 2005

Specification Update 7

Preface

This document is an update to the specifications contained in the documents listed in the following Affected Documents/Related Documents table. It is a compilation of device and document errata and specification clarifications and changes, and is intended for hardware system manufacturers and for software developers of applications, operating system, and tools.

Information types defined in the Nomenclature section of this document are consolidated into this update document and are no longer published in other documents. This document may also contain information that has not been previously published.

Affected Documents

Preface

Pentium® III Processor for the SC242 at 450 MHz to 1.13 GHz datasheet

Pentium® III Processor for the PGA370 Socket up to 1.13 GHz datasheet

Related Documents

Intel Architecture Software Developer’s Manual, Volumes 1, 2a, 2b, 3a and 3b

Nomenclature

Errata are design defects or errors. Errata may cause the Pentium III behavior to

deviate from published specifications. Hardware and software designed to be used with any given stepping must assume that all errata documented for that stepping are present on all devices.

Specification Changes are modifications to the current published specifications. These changes will be incorporated in the next release of the specifications.

Document Title Document

Number/Location

244452

245264

Number/Location

253665, 253666, 253667, 253668,

253669

Specification Clarifications describe a specification in greater detail or further highlight a specification’s impact to a complex design situation. These clarifications will be incorporated in the next release of the specifications.

8 Specification Update

Preface

Documentation Changes include typos, errors, or omissions from the current published specifications. These changes will be incorporated in the next release of the specifications.

Note: Errata remain in the specification update throughout the product’s lifecycle, or until a

particular stepping is no longer commercially available. Under these circumstances, errata removed from the specification update are archived and available upon request. Specification changes, specification clarifications and documentation changes are removed from the specification update when the appropriate changes are made to the appropriate product specification or user documentation (datasheets, manuals, etc.).

Specification Update 9

Summary Tables of Changes

The following table indicates the Specification Changes, Errata, Specification Clarifications or Documentation Changes, which apply to the listed MCH steppings. Intel intends to fix some of the errata in a future stepping of the component, and to account for the other outstanding issues through documentation or Specification Changes as noted. This table uses the following notations:

Codes Used in Summary Table

Stepping

X: Erratum, Specification Change or Clarification that applies

to this stepping.

Status

Row

(No mark) or (Blank Box): This erratum is fixed in listed stepping or specification

change does not apply to listed stepping.

Doc: Document change or update that will be implemented.

PlanFix: This erratum may be fixed in a future stepping of the

product.

Fixed: This erratum has been previously fixed.

NoFix: There are no plans to fix this erratum.

Shaded: This item is either new or modified from the previous

version of the document.

Each Specification Update item is prefixed with a capital letter to distinguish the product. The key below details the letters that are used in Intel’s microprocessor Specification Updates:

A = Dual-Core Intel® Xeon® processor 7000 sequence

C = Intel® Celeron® processor

10 Specification Update

Summary Tables of Changes

D = Dual-Core Intel® Xeon® processor 2.80 GHz

E = Intel® Pentium® III processor

F = Intel® Pentium® processor Extreme Edition and Intel® Pentium® D

processor

I = Dual-Core Intel® Xeon® processor 5000 series

J = 64-bit Intel® Xeon® processor MP with 1MB L2 cache

K = Mobile Intel® Pentium® III processor

L = Intel® Celeron® D processor

M = Mobile Intel® Celeron® processor

N = Intel® Pentium® 4 processor

O = Intel® Xeon® processor MP

P = Intel ® Xeon® processor

Q = Mobile Intel® Pentium® 4 processor supporting Hyper-Threading technology

on 90-nm process technology

R = Intel® Pentium® 4 processor on 90 nm process

S = 64-bit Intel® Xeon® processor with 800 MHz system bus (1 MB and 2 MB L2

cache versions)

T = Mobile Intel® Pentium® 4 processor-M

U = 64-bit Intel® Xeon® processor MP with up to 8MB L3 cache

V = Mobile Intel® Celeron® processor on .13 micron process in Micro-FCPGA

package

W= Intel® Celeron® M processor

X = Intel® Pentium® M processor on 90nm process with 2-MB L2 cache and

Intel® processor A100 and A110 with 512-KB L2 cache

Y = Intel® Pentium® M processor

Z = Mobile Intel® Pentium® 4 processor with 533 MHz system bus

AA = Intel® Pentium® D processor 900 sequence and Intel® Pentium® processor

Extreme Edition 955, 965

AB = Intel® Pentium® 4 processor 6x1 sequence

AC = Intel(R) Celeron(R) processor in 478 pin package

AD = Intel(R) Celeron(R) D processor on 65nm process

AE = Intel® Core™ Duo processor and Intel® Core™ Solo processor on 65nm

process

AF = Dual-Core Intel® Xeon® processor LV

AG = Dual-Core Intel® Xeon® processor 5100 series

AH = Intel® Core™2 Duo/Solo processor for Intel® Centrino® Duo processor

technology

Specification Update 11

Summary Tables of Changes

AI = Intel® Core™2 Extreme processor X6800 and Intel® Core™2 Duo desktop

processor E6000 and E4000 sequence

AJ = Quad-Core Intel® Xeon® processor 5300 series

AK = Intel® Core™2 Extreme quad-core processor QX6000 sequence and Intel®

Core™2 Quad processor Q6000 sequence

AL = Dual-Core Intel® Xeon® processor 7100 series

AM = Intel® Celeron® processor 400 sequence

AN = Intel® Pentium® dual-core processor

AO = Quad-Core Intel® Xeon® processor 3200 series

AP = Dual-Core Intel® Xeon® processor 3000 series

AQ = Intel® Pentium® dual-core desktop processor E2000 sequence

AR = Intel® Celeron® processor 500 series

AS = Intel® Xeon® processor 7200, 7300 series

AT = Intel® Celeron® processor 200 series

AU = Intel® Celeron® Dual Core processor T1400

AV = Intel® Core™2 Extreme processor QX9650 and Intel® Core™2 Quad

processor Q9000 series

AW = Intel® Core™ 2 Duo processor E8000 series

AX = Quad-Core Intel® Xeon® processor 5400 series

AY= Dual-Core Intel® Xeon® processor 5200 series

AZ = Intel® Core™2 Duo Processor and Intel® Core™2 Extreme Processor on 45-

nm Process

AAA = Quad-Core Intel® Xeon® processor 3300 series

AAB = Dual-Core Intel® Xeon® E3110 Processor

AAC = Intel® Celeron® dual-core processor E1000 series

AAD = Intel® Core™2 Extreme Processor QX9775Δ

AAE = Intel® Atom™ processor Z5xx series

AAF = Intel® Atom™ processor 200 series

AAG = Intel® Atom™ processor N series

The Specification Updates for the Pentium processor, Pentium

Pro processor, and

other Intel products do not use this convention.

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

E1 X X X X X X X X NoFix

12 Specification Update

FP data operand pointer may be incorrectly calculated after FP access which wraps 64-Kbyte boundary in 16-bit code

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

E2 X X X X X X X X NoFix

E3 X X X X X X X X NoFix

E4 X X X X X X X X NoFix

E5 X X X X X X X X NoFix

E6 X X X X X X X X NoFix

E7 X X X X X X X X NoFix BTM for SMI will contain incorrect FROM EIP

E8 X X X X X X X X NoFix

E9 X X X X X X X X NoFix

E10 X X Fixed

E11 X X Fixed

E12 X X X X X X X X NoFix

E13 X X X X X X X X NoFix

E14 X X X X X X X X NoFix LBER may be corrupted after some events

E15 X X X X X X X X NoFix

E16 X X X X X X X X NoFix

E17 X X X X X X X X NoFix

E18 X X X X X X X X NoFix

E19 X X Fixed

E20 X X X X X X X X NoFix

E21 X X X X X X X X NoFix

E22 X X X X X X X X NoFix

E23 X X X X X X X X NoFix

Differences exist in debug exception reporting

FLUSH# servicing delayed while waiting for STARTUP_IPI in 2-way MP systems

Code fetch matching disabled debug register may cause debug exception

Double ECC error on read may result in BINIT#

FP inexact-result exception flag may not be set

I/O restart in SMM may fail after simultaneous MCE

Branch traps do not function if BTMs are also enabled

Checker BIST failure in FRC mode not signaled

BINIT# assertion causes FRCERR assertion in FRC mode

Machine check exception handler may not always execute successfully

MCE due to L2 parity error gives L1 MCACOD.LL

BTMs may be corrupted during simultaneous L1 cache line replacement

EFLAGS discrepancy on a page fault after a multiprocessor TLB shootdown

Near CALL to ESP creates unexpected EIP address

Memory type undefined for nonmemory operations

Infinite snoop stall during L2 initialization of MP systems

FP data operand pointer may not be zero after power on or Reset

MOVD following zeroing instruction can cause incorrect result

Premature execution of a load operation prior to exception handler invocation

Read portion of RMW instruction may execute twice

Specification Update 13

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

MC2_STATUS MSR has model-specific error

E24 X X X X X X X X NoFix

E25 X X X X X X X X NoFix

E26 X X X X X X X X NoFix

E27 X X X X X X X X NoFix

E28 X X X X X X X X NoFix

E29 X X X X X X X X NoFix

E30 X X X X X X X X NoFix

E31 X X X X X X X X NoFix

E32 X X Fixed

E33 X X X X X X X X NoFix

E34 X Fixed

E35 X Fixed Transmission error on cache read

E36 X X X X X X X X NoFix

E37 X X X X X X X X NoFix

E38 X X Fixed

E39 X X X X X X X X NoFix

E40 X Fixed

E41 X X X X X X X X NoFix System bus ECC not functional with 2:1 ratio

E42 X X X X X X X X NoFix

E43 X X X X X X X X NoFix

E44 X X X X Fixed

code and machine check architecture error code reversed

Mixed cacheability of lock variables is problematic in MP systems

MOV with debug register causes debug exception

Upper four PAT entries not usable with Mode B or Mode C paging

Data breakpoint exception in a displacement relative near call may corrupt EIP

RDMSR and WRMSR to invalid MSR may not cause GP fault

SYSENTER/SYSEXIT instructions can implicitly load null segment selector to SS and CS registers

PRELOAD followed by EXTEST does not load boundary scan data

Far jump to new TSS with D-bit cleared may cause system hang

INT 1 instruction handler execution could generate a debug exception

COMISS/UCOMISS may not update EFLAGS under certain conditions

Potential loss of data coherency during MP data ownership transfer

Misaligned Locked access to APIC space results in hang

Floating-point exception signal may be deferred

Memory ordering based synchronization may cause a livelock condition in mp systems

System bus address parity generator may report false AERR#

Processor may assert DRDY# on a write with no data

GP# fault on WRMSR to ROB_CR_BKUPTMPDR6

Machine check exception may occur due to improper line eviction in the IFU

14 Specification Update

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

E45 X X X Fixed

E46 X X X X X Fixed Snoop request may cause DBSY# hang

E47 X X X X X X X X NoFix

E48 X X X Fixed

E49 X X X X X X X X NoFix

E50 X Fixed

E51 X Fixed Noise sensitivity issue on processor SMI# pin

E52 X Fixed

E53 X Fixed

E54 X Fixed IFU/DCU deadlock may cause system hang

E55 X Fixed

E56 X X Fixed

E57 X X X X Fixed

E58 X X X X Fixed

E59 X X X X X X X X NoFix

E60 X X X X X X X X NoFix

E61 X Fixed

E62 X X X X X X Fixed

E63 X Fixed

E64 X X Fixed

E65 X X X X X X X X NoFix

Performance counters include Streaming SIMD Extensions L1 prefetch

Lower bits of SMRAM SMBASE register cannot be written with an ITP

Task Switch May Cause Wrong PTE and PDE Access Bit to be Set

Unsynchronized Cross-Modifying code operations can cause unexpected instruction execution results

Processor will erroneously report a BIST failure

Limitation on cache line ECC detection and correction

L2_LD and L2_M_LINES_OUTM performancemonitoring counters do not work

L2_DBUS_BUSY performance monitoring counter will not count writes

Incorrect sign may occur on X87 result due to indefinite QNaN result from streaming SIMD extensions multiply

Deadlock may occur due to illegalinstruction/page-miss combination

MASKMOVQ instruction interaction with string operation may cause deadlock

MOVD, CVTSI2SS, or PINSRW Following Zeroing Instruction Can Cause Incorrect Result

FLUSH# assertion following STPCLK# may prevent CPU clocks from stopping

Intermittent failure to assert ADS# during processor power-on

Floating-point exception condition may be deferred

THERMTRIP# may not be asserted as specified

Cache line reads may result in eviction of invalid data.

Snoop probe during FLUSH# could cause L2 to be left in shared state

Specification Update 15

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

E66 X X X X Fixed Livelock may occur due to IFU line eviction

E67 X X X X Fixed

E68 X X X X X X X X NoFix INIT does not clear global entries in the TLB

E69

E70

E71 X X X X X X X X NoFix

E72 X X X X X X X X NoFix

E73 X X X X X X X X NoFix

E74 X X X X X X X X NoFix

E75 X X X X X X X X NoFix

E76 X Fixed

E77 X X X X X X X X PlanFix

E78 X X X X X X X NoFix Incorrect assertion of THERMTRIP# Signal

E79 X NoFix

E80 X X X X X X X1 Fixed

E81 X X X X X X X X NoFix

E82 X X X X X X X X NoFix

E83 X X X X X X X X NoFix IFU/BSU Deadlock May Cause System Hang

E84

X X X X X X X X NoFix VM bit will be cleared on a double fault

X X X X X X X X NoFix Memory aliasing with inconsistent A and D

X X X X X X X X NoFix REP MOVS Operation in Fast string Mode

Selector for the LTR/LLDT register may get corrupted

handler

bits may cause processor deadlock

Use of memory aliasing with inconsistent memory type may cause system hang

Processor may report invalid TSS fault instead of Double fault during mode C paging

Machine check exception may occur when interleaving code between different memory types

Wrong ESP register values during a fault in VM86 mode

APIC ICR write may cause interrupt not to be sent when ICR delivery bit pending

High temperature and low supply voltage operation may result in incorrect processor operation

During Boundary Scan, BCLK not Sampled High When SLP# is Asserted Low

Processor might not exit Sleep State properly upon de-assertion of CPUSLP# signal

The Instruction Fetch Unit (IFU) may fetch instructions based upon stale CR3 data after a write to CR3 register

Under Some Complex Conditions, the Instructions in the Shadow of a JMP FAR may be Unintentionally Executed and Retired

Processor Does not Flag #GP on Non-zero Write to Certain MSRs

Continues in that Mode When Crossing into a Page with a Different Memory Type

16 Specification Update

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

X X X X X X X X NoFix The FXSAVE, STOS, MOVS Instructions May

E85

E86 X X X X X X X X NoFix

E87 X X X X X X X X NoFix

E88 X X X X X X X X NoFix

E89 X X X X X X X X NoFix

E90 X X X X X X X X NoFix

E91 X X X X X X X X NoFix

E92 X X X X X X X X NoFix

E93 X X X X X X X X NoFix

E94 X X X X X X X X NoFix

E95 X X X X X X X X NoFix

E96 X X X X X X X X NoFix

E97 X X X X X X X X NoFix

E98 X X X X X X X X NoFix

Cause a Store Ordering Violation When Data Crosses a Page with a UC Memory Type

POPF and POPFD Instructions that Set the Trap Flag Bit May Cause Unpredictable Processor Behavior

Code Segment Limit Violation May Occur on 4 Gbyte Limit Check

FST Instruction with Numeric and Null Segment Exceptions May Cause General Protection Faults to be Missed and FP Linear Address (FLA) Mismatch

Code Segment is Wrong on SMM Handler when SMBASE is not Aligned

Page with PAT (Page Attribute Table) Set to USWC (Uncacheable Speculative Write Combine) While Associated MTRR (Memory Type Range Register) is UC (Uncacheable) May Consolidate to UC

Under Certain Conditions LTR (Load Task Register) Instruction May Result in System Hang

Loading from Memory Type USWC (Uncacheable Speculative Write Combine) May Get Its Data Internally Forwarded From a Previous Pending Store

FXSAVE after FNINIT Without an Intervening FP (Floating Point) Instruction May Save Uninitialized Values for FDP (x87 FPU Instruction Operand (Data) Pointer Offset) and FDS (x87 FPU Instruction Operand (Data) Pointer Selector)

FSTP (Floating Point Store) Instruction Under Certain Conditions May Result In Erroneously Setting a Valid Bit on an FP (Floating Point) Stack Register

Invalid Entries in Page-Directory-PointerTable-Register (PDPTR) May Cause General Protection (#GP) Exception if the Reserved Bits are Set to One

Writing the Local Vector Table (LVT) when an Interrupt is Pending May Cause an Unexpected Interrupt

The Processor May Report a #TS Instead of a #GP Fault

A Write to an APIC Register Sometimes May Appear to Have Not Occurred

Specification Update 17

Summary Tables of Changes

NO. kB0 kC0 cA2 cB0 cC0 cD0 tA1 tB1 Plans ERRATA

Using 2M/4M Pages When A20M# Is

E99 X X X X X X X X NoFix

Asserted May Result in Incorrect Address Translations

E100 X X X X X X X X NoFix

E101 X X X X X X X X NoFix INIT Does Not Clear Global Entries in the TLB

E102 X X X X X X X X NoFix

E103 X X X X X X X X NoFix

E104 X X X X X X X X NoFix

E105 X X X X X X X X NoFix

E106 X X X X X X X X NoFix

E107 X X X X X X X X NoFix

E108 X X X X X X X X NoFix

E109 X X X X X X X X NoFix

Values for LBR/BTS/BTM will be Incorrect after an Exit from SMM

REP MOVS/STOS Executing with Fast Strings Enabled and Crossing Page Boundaries with Inconsistent Memory Types may use an Incorrect Data Size or Lead to MemoryOrdering Violations

The BS Flag in DR6 May be Set for NonSingle-Step #DB Exception

Fault on ENTER Instruction May Result in Unexpected Values on Stack Frame

Unaligned Accesses to Paging Structures May Cause the Processor to Hang

INVLPG Operation for Large (2M/4M) Pages May be Incomplete under Certain Conditions

Page Access Bit May be Set Prior to Signaling a Code Segment Limit Fault

EFLAGS, CR0, CR4 and the EXF4 Signal May be Incorrect after Shutdown

Performance Monitoring Event FP_MMX_TRANS_TO_MMX May Not Count Some Transitions

* Fix will be only on Pentium® III processors with CPUID=068xh and not CPUID=067xh

Notes:

For these steppings, this erratum may be worked around in BIOS.

Number SPECIFICATION CHANGES

There are no Specification Changes in this revision of the Specification Update.

Number SPECIFICATION CLARIFICATIONS

E1 Specification clarification with respect to time stamp counter.

18 Specification Update

Summary Tables of Changes

Number DOCUMENTATION CHANGES

There are no Documentation Chnages in this revision of the Specification Update

Specification Update 19

Identification Information

Identification Informati o n

Component Iden tification via Pr ogramming Interface

The Pentium® III processor can be identified by the following values:

Family1 Model2 Brand ID3

0110 0111 00h = Not Supported

0110 1000 02h = "Intel® Pentium® III

Processor"

NOTES:

1. The Family corresponds to bits [11:8] of the EDX register after RESET, bits [11:8] of

the EAX register after the CPUID instruction is executed with a 1 in the EAX register, and the generation field of the Device ID register accessible through Boundary Scan.

2. The Model corresponds to bits [7:4] of the EDX register after RESET, bits [7:4] of the

EAX register after the CPUID instruction is executed with a 1 in the EAX register, and the model field of the Device ID register accessible through Boundary Scan.

3. The Brand ID corresponds to bits [7:0] of the EBX register after the CPUID instruction is

executed with a 1 in the EAX register.

The Pentium III processor’s second level (L2) cache size can be determined by the following register contents:

512-Kbyte Unified L2 Cache1 43h

256-Kbyte 8 way set associative

32byte line size, L2 Cache

512-Kbyte 8 way set associative

32byte line size, L2 Cache

NOTES:

1. For the Pentium III processor, the unified L2 cache size corresponds to a token in the

EDX register after the CPUID instruction is executed with a 2 in the EAX register. Other Intel microprocessor models or families may move this information to other bit positions or otherwise reformat the result returned by this instruction; generic code should parse the resulting token stream according to the definition of the CPUID instruction.

82h

83h

20 Specification Update

Identification Information

Mixed Step pings i n DP Systems

Intel Corporation fully supports mixed steppings of Pentium III processors. The following list and processor matrix describes the requirements to support mixed steppings:

• Mixed steppings are only supported with processors that have identical family and model

number as indicated by the CPUID instruction.

• While Intel has done nothing to specifically prevent processors operating at differing

frequencies from functioning within a multiprocessor system, there may be uncharacterized errata that exist in such configurations. Intel does not support such configurations. In mixed stepping systems, all processors must operate at identical frequencies (i.e., the highest frequency rating commonly supported by all processors).

• While there are no known issues associated with the mixing of processors with differing

cache sizes in a dual processor system, and Intel has done nothing to specifically prevent such system configurations from operating, Intel does not support such configurations since there may be uncharacterized errata that exist. In dual processor systems, all processors must be of the same cache size.

• While Intel believes that certain customers may wish to perform validation of system

configurations with mixed frequency or cache sizes, and that those efforts are an acceptable option to our customers, customers would be fully responsible for the validation of such configurations.

• The workarounds identified in this and following specification updates must be properly

applied to each processor in the system. Certain errata are specific to the dual processor environment and are identified in the Mixed Stepping Processor Matrix found at the end of this section. Errata for all processor steppings will affect system performance if not properly

worked around. Also see the “Pentium® III Processor Identification and Package Information” table for additional details on which processors are affected by specific errata.

• In dual processor systems, the processor with the lowest feature-set, as determined by the

CPUID Feature Bytes, must be the Bootstrap Processor (BSP). In the event of a tie in feature-set, the tie should be resolved by selecting the BSP as the processor with the lowest stepping as determined by the CPUID instruction.

In the following processor matrix a number indicates that a known issue has been identified as listed in the table following the matrix. A dual processor system using mixed processor steppings must assure that errata are addressed appropriately for each processor.

Specification Update 21

MHz

600E

MHz

550E

600E

MHz

600E

MHz

1 GHz

Identification Information

DP Platform Population Matrix for t he Pentium® III Processor with 100-MHz System Bus in the SECC and SECC2 Packages

Pentium® III

Processor

Stepping

450-MHz kB0 NI X NI X X X X X X X X X X X X X X X X X X X X X X 500-MHz kB0 X NI X NI X X X X X X X X X X X X X X X X X X X X X 450-MHz kC0 NI X NI X X X X X X X X X X X X X X X X X X X X X X 500-MHz kC0 X NI X NI X X X X X X X X X X X X X X X X X X X X X 550-MHz kC0 X X X X NI X X X X X X X X X X X X X X X X X X X X 600-MHz kC0 X X X X X NI X X X X X X X X X X X X X X X X X X X 600E-MHz cA2 X X X X X X NI X X X X X NI X X X X X NI X X X X X X 650-MHz cA2 X X X X X X X NI X X X X X NI X X X X X NI X X X X X 700-MHz cA2 X X X X X X X X NI X X X X X NI X X X X X NI X X X X 750-MHz cA2 X X X X X X X X X NI X X X X X NI X X X X X NI X X X 800 MHz cA2 X X X X X X X X X X NI X X X X X NI X X X X X NI X X 550E MHz cB0 X X X X X X X X X X X NI X X X X X X X X X X X X X 600E MHz cB0 X X X X X X NI X X X X X NI X X X X X NI X X X X X X 650 MHz cB0 X X X X X X X NI X X X X X NI X X X X X NI X X X X X 700 MHz cB0 X X X X X X X X NI X X X X X NI X X X X X NI X X X X 750 MHz cB0 X X X X X X X X X NI X X X X X NI X X X X X NI X X X 800 MHz cB0 X X X X X X X X X X NI X X X X X NI X X X X X NI X X 850 MHz cB0 X X X X X X X X X X X X X X X X X NI X X X X X NI X 600E-MHz cC0 X X X X X X NI X X X X X NI X X X X X NI X X X X X X 650 MHz cC0 X X X X X X X NI X X X X X NI X X X X X NI X X X X X 700 MHz cC0 X X X X X X X X NI X X X X X NI X X X X X NI X X X X 750-MHz cC0 X X X X X X X X X NI X X X X X NI X X X X X NI X X X 800 MHz cC0 X X X X X X X X X X NI X X X X X NI X X X X X NI X X 850 MHz cC0 X X X X X X X X X X X X X X X X X NI X X X X X NI X 1 GHz cC0 X X X X X X X X X X X X X X X X X X X X X X X X NI

450

kB0

500

kB0

450

kC0

500

kC0

550

kC0

600

kC0

cA2

650

cA2

700

cA2

750

cA2

800

cA2

MHz

cB0

650

cB0

700

cB0

750

cB0

800

cB0

850

cB0

cC0

650

cC0

700

cC0

750

cC0

800

cC0

850

cC0

NOTES:

1. X = Mixing processors at different frequencies is not supported. This

stepping/frequency not supported in DP.

2. NI = Currently no known issues associated with mixing these steppings.

22 Specification Update

Identification Information

DP Platform Population Matrix for the Pentium® III Processor with 133-MHz System Bus in the SECC and SECC2 Package from 533MHz to 733MHz

Pentium® III

Processor

Stepping

533B-MHz kB0 NI NI X 533B-MHz kC0 NI NI X 600B-MHz kC0 X X NI 533EB-MHz cA2 X X X 600EB-MHz cA2 X X X 667-MHz cA2 X X X 733-MHz cA2 X X X 533EB MHz cB0 X X X 600EB MHz cB0 X X X 667 MHz cB0 X X X 733 MHz cB0 X X X 600EB-MHz cC0 X X X 667-MHz cC0 X X X 733-MHz cC0 X X X

533B

MHz

kB0

533B

MHz

kC0

533E

600E

667

733

533E

600E

667

600B

B MHz

MHz

B MHz

MHz

cA2

kC0

X X X X X X X X X X X

NI X X X NI X X X X X X

X NI X X X NI X X NI X X X X NI X X X X X NI

NI X X X NI

X NI X X NI X X X NI X X X NI X X NI X NI X X X NI X X NI X X X X NI X X X NI X X NI X X X X NI X X X NI X X NI

cA2

X X

B MHz

cB0

X X

X X X

X NI

X NI X X NI X

733

MHz

cB0

NI X X NI X

X NI X X NI

X X

B MHz

cB0

600E

667

733

MHz

cC0

MHz

cC0

X X X

NI X X

NOTES:

1. X = Mixing processors at different frequencies is not supported. This

stepping/frequency is not supported in Dual Processor.

2. NI = Currently no known issues associated with mixing these steppings.

DP Platform Population Matrix for the Pentium® III Processor with 133-MHz System Bus in the SECC and SECC2 Package from 800MHz to 1.13GHz

Pentium® III Processor

Stepping

800EB-MHz cA2 800EB MHz cB0 866 MHz cB0 933 MHz cB0 1B GHz cB0 800EB MHz cC0 866 MHz cC0 933 MHz cC0 1B GHz cC0

1.13 GHz cC0

800EB

MHz cA2

NI NI X X X NI X X X X NI NI X X X NI X X X X

NI NI X X X NI X X X X

NOTES:

1. X = Mixing processors at different frequencies is not supported. This

stepping/frequency is not supported in Dual Processor.

2. NI = Currently no known issues associated with mixing these steppings.

800EB

MHz cB0

X X NI X X X NI X X X X X X NI X X X NI X X

X X X X X X X X X X

X X NI X X X NI X X X X X X NI X X X NI X X X X X X X X X X NI X X X X X X X X X X X

866 MHz

cB0

933 MHz cB0,

1B GHz

cB0

800EB

MHz cC0

866 MHz cC0,

933 MHz

cC0,

GHz

cC0

1.13

GHz cC0

Specification Update 23

Identification Information

DP Platform Population Matrix for the Pentium® III Processor with 100MHz System Bus in the FC-PGA370 Package from 500 MHz to 650 MHz

Pentium® III

Processor

Stepping

500E-MHz cB0 NI X X X

500E MHz

cB0

550E MHz

cB0

600E MHz

cB0

650 MHz

cB0

600E MHz

cC0

650 MHz

cC0

600E MHz

cD0

X X X

550E-MHz cB0 X NI X X

X X X

600E-MHz cB0 X X NI X

X X NI

650-MHz cB0 X X X NI

X X X

600E-MHz cC0 X X X X NI

X NI

650-MHz cC0 X X X X

600E-MHz cD0 X X X X

NI X

DP Platform Population Matrix for the Pentium® III Processor with 100-MHz System Bus in the FC-PGA 370 Pin Package from 700 MHz to 1.10 GHz

Pentium® III

Processor

Stepping

700-MHz cB0 NI X X X NI X X X X NI X X X X X X 750-MHz cB0 X NI X X X NI X X X X NI X X X X X 800-MHz cB0 X X NI X X X NI X X X X NI X X X X 850-MHz cB0 X X X NI X X X NI X X X X NI X X X 700-MHz cC0 NI X X X NI X X X X NI X X X X X X 750-MHz cC0 X NI X X X NI X X X X NI X X X X X 800-MHz cC0 X X NI X X X NI X X X X NI X X X X 850-MHz cC0 X X X NI X X X NI X X X X NI X X X 900-MHz cC0 X X X X X X X X NI X X X X NI X X 700-MHz cD0 NI X X X NI X X X X NI X X X X X X 750-MHz cD0 X NI X X X NI X X X X NI X X X X X 800-MHz cD0 X X NI X X X NI X X X X NI X X X X 850-MHz cD0 X X X NI X X X NI X X X X NI X X X 900-MHz cD0 X X X X X X X X NI X X X X NI X X 1 GHz cD0 X X X X X X X X X X X X X X NI X

1.10GHz cD0 X X X X X X X X X X X X X X X NI

700

750

800

850

700

750

800

850

900

700

750

MHz

cB0

cC0

NOTES:

1. X = Mixing processors at different frequencies is not supported.

2. NI = Currently no known issues associated with mixing these steppings.

3. TBD = No issues are expected, however further investigation is required to fully

validate this DP solution.

cC0

MHz

cD0

MHz

cD0

800

MHz

cD0

MHz

850

cD0

900

MHz

cD0

GHz

cD0

1.10 GHz cD0

24 Specification Update

800EB

MHz

Identification Information

DP Platform Population Matrix for the Pentium® III Processor with 133MHz System Bus in the FC - PGA370 Packag e from 533 MHz to 800 MHz

Pentium® III

Processor

Stepping

533EB-MHz cB0 NI X X X X X X X X X X X X 600EB-MHz cB0 X NI X X X X X NI X X X X X 667-MHz cB0 X X NI X X X X X NI X X X X 733-MHz cB0 X X X NI X X X X X NI X NI X 800EB-MHz cB0 X X X X NI X X X X X NI X NI 866-MHz cB0 X X X X X NI X X X X X X X 933-MHz cB0 X X X X X X NI X X X X X X 600EB-MHz cC0 X NI X X X X X NI X X X X X 667-MHz cC0 X X NI X X X X X NI X X X X 733-MHz cC0 X X X NI X X X X X NI X X X 800EB-MHz cC0 X X X X NI X X X X X NI X X 733-MHz cD0 X X X X X X X X NI X NI X

800EB-MHz cD0

667

733

866

933

667

733

533EB

600E

MHz

cB0

X X X X NI X X X X X NI X NI

MHz

cB0

MHz

cB0

MHz

cB0

800E

MHz

cB0

MHz

cB0

MHz

cB0

600E

MHz

cC0

MHz

cC0

MHz

cC0

800E

B MHz cC0

733

MHz

cD0

NOTES:

1. X = Mixing processors at different frequencies is not supported.

2. NI = Currently no known issues associated with mixing these steppings.

3. TBD = No issues are expected, however further investigation is required to fully

validate this DP solution.

DP Platform Population Matrix for the Pentium® III Processor with 133-MHz System Bus in the FC-PGA370 Package from 866 MHz to 1 GHz

Pentium® III

Processor

Stepping

866-MHz cB0 933-MHz cB0 866-MHz cC0 933-MHz cC0 1B-GHz cC0 866-MHz cD0 933-MHz cD0 1B-GHz cD0

NOTES:

1. X = Mixing processors at different frequencies is not supported.

2. NI = Currently no known issues associated with mixing these steppings.

3. TBD = No issues are expected, however further investigation is required to fully

866 MHz

cB0

NI X NI X X NI X X

validate this DP solution.

933 MHz

cB0

X NI X NI X X NI X

X NI X NI X X NI X X X X X NI X X NI

866 MHz

cC0

933 MHz

cC0

1B GHz

cC0

866 MHz

cD0

933 MHz

cD0

1B GHz

cD0

Specification Update 25

Identification Information

DP Platform Population Matrix for the Pentium®III Processor with 133-MHz System Bus in the FC-PGA2 Package from 866 MH z t o 1.4 GH z and uF CBGA2 Package for 800 MHz to 933 MHz

Pentium® III

Processor

Stepping

866-MHz cD0

933-MHz cD0

1B-GHz cD0 X X NI X X X X X X X X X X X

1.13-GHz cD0

800-MHz tA1 (uFCBGA2)

800-MHz tB1 (uFCBGA2)

933-MHz tB1 (uFCBGA2)

1-GHz tA1 X X X X X X X NI X X X X X X

1.13-GHz tA1

1.26-GHz tA1

1.4-GHz tA1 X X X X X X X X X X X X NI NI

1.13-GHz tB1

1.26-GHz tB1

1.4-GHz tB1 X X X X X X X X X X X X NI NI

866

MHz

cD0

933

MHz

cD0

NI X X X X X X X X X X X X X

X NI X X X X X X X X X X X X

X X X NI X X X X X X X X X X

X X X X NI NI X X X X X X X X

X X X X X X NI X X X X X X X

X X X X X X X X NI NI X X X X

X X X X X X X X X X NI NI X X

X X X X X X X X NI NI X X X X

X X X X X X X X X X NI NI X X

GHz

cD0

1.13

GHz

cD0

800

MHz

tA1

800

MHz

tB1

933

MHz

tB1

GHz

tA1

1.13 GHz

tA1

1.13 GHz

tB1

1.26 GHz

tA1

1.26 GHz

tB1

1.4

GHz

tA1

1.4

GHz

tB1

NOTES:

1. X = Mixing processors at different frequencies is not supported.

2. NI = Currently no known issues associated with mixing these steppings.

3. TBD = No issues are expected, however further investigation is required to fully

validate this DP solution.

26 Specification Update

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

SL364 kB0 0672h 450/100 512 T6P-e/A0 ECC D SECC2† 1, 2, 4

SL365 kB0 0672h 500/100 512 T6P-e/A0 ECC D SECC2† 1, 2, 4, 8

SL3CC kB0 0672h 450/100 512 T6P-e/A0 ECC D SECC2† 1, 2, 3, 4

SL3CD kB0 0672h 500/100 512 T6P-e/A0 ECC D SECC2† 1, 2, 3, 4

SL38E kB0 0672h 450/100 512 T6P-e/A0 ECC D S.E.C.C 1, 2, 4

SL38F kB0 0672h 500/100 512 T6P-e/A0 ECC D S.E.C.C 1, 2, 4

SL35D kC0 0673h 450/100 512 T6P-e/A0 ECC E SECC2† 1, 4

SL37C kC0 0673h 450/100 512 T6P-e/A0 ECC E SECC2† 1, 3, 4

SL35E kC0 0673h 500/100 512 T6P-e/A0 ECC E SECC2† 1, 4

SL37D kC0 0673h 500/100 512 T6P-e/A0 ECC E SECC2† 1, 3, 4

SL3F7 kC0 0673h 550/100 512 T6P-e/A0 ECC E SECC2† 1, 4

SL3FJ kC0 0673h 550/100 512 T6P-e/A0 ECC E SECC2† 1, 3, 4

SL3BN kC0 0673h 533B/133 512 T6P-e/A0 ECC E SECC2† 1, 4, 10

SL3E9 kC0 0673h 533B/133 512 T6P-e/A0 ECC E SECC2† 1, 3, 4, 10

SL3JM kC0 0673h 600/100 512 T6P-e/A0 ECC E SECC2† 1, 4

SL3JT kC0 0673h 600/100 512 T6P-e/A0 ECC E SECC2† 1, 3, 4

SL3JP kC0 0673h 600B/133 512 T6p-e/A0 ECC E SECC2† 1, 4, 10

SL3JU kC0 0673h 600B/133 512 T6P-e/A0 ECC E SECC2† 1, 3, 4, 10

SL3Q9 cA2 0681h 500E/100 256 N/A ECC B FC-PGA

SL3R2 cA2 0681h 500E/100 256 N/A ECC B FC-PGA

SL3VF cA2 0681h 533EB/133 256 N/A ECC B FC-PGA

SL3VA cA2 0681h 533EB/133 256 N/A ECC B FC-PGA

SL3QA cA2 0681h 550E/100 256 N/A ECC B FC-PGA

SL3R3 cA2 0681h 550E/100 256 N/A ECC B FC-PGA

SL3VH cA2 0681h 600E/100 256 N/A ECC B FC-PGA

SL3NL cA2 0681h 600E/100 256 N/A ECC B FC-PGA

Core

Stepping CPUID

Speed

(MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

9, 10

(370 pin)

7, 9, 10

(370 pin)

9, 10

(370 pin)

7, 9, 10

(370 pin)

9, 10

(370 pin)

7, 9, 10

(370 pin)

9, 10

(370 pin)

7, 9, 10

(370 pin)

Specification Update 27

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL3VG cA2 0681h 600EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3VB cA2 0681h 600EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3VJ cA2 0681h 650/100 256 N/A ECC B FC-PGA

(370 pin)

SL3NM cA2 0681h 650/100 256 N/A ECC B FC-PGA

(370 pin)

SL3VK cA2 0681h 667/133 256 N/A ECC B FC-PGA

(370 pin)

SL3T2 cA2 0681h 667/133 256 N/A ECC B FC-PGA

(370 pin)

SL3VL cA2 0681h 700/100 256 N/A ECC B FC-PGA

(370 pin)

SL3T3 cA2 0681h 700/100 256 N/A ECC B FC-PGA

(370 pin)

SL3VM cA2 0681h 733/133 256 N/A ECC B FC-PGA

(370 pin)

SL3T4 cA2 0681h 733/133 256 N/A ECC B FC-PGA

(370 pin)

SL3VN cA2 0681h 750/100 256 N/A ECC B FC-PGA

(370 pin)

SL3VC cA2 0681h 750/100 256 N/A ECC B FC-PGA

(370 pin)

SL3WB cA2 0681h 800EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3VE cA2 0681h 800EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3X4 cA2 0681h 800/100 256 N/A ECC B FC-PGA

(370 pin)

SL3VD cA2 0681h 800/100 256 N/A ECC B FC-PGA

(370 pin)

SL444 cB0 0683h 500E/100 256 N/A ECC B FC-PGA

(370 pin)

SL446 cB0 0683h 500E/100 256 N/A ECC B FC-PGA

(370 pin)

SL45R cB0 0683h 500E/100 256 N/A ECC B FC-PGA

(370 pin)

SL3XS cB0 0683h 533EB/133 256 N/A ECC B FC-PGA

(370 pin)

9, 10

7, 9, 10

7, 9

9, 10

7, 9, 10

9, 10

7, 9, 10

10, 7

28 Specification Update

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL45S cB0 0683h 533EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL44G cB0 0683h 550E/100 256 N/A ECC B FC-PGA

(370 pin)

SL45T cB0 0683h 550E/100 256 N/A ECC B FC-PGA

(370 pin)

SL3XT cB0 0683h 600EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL45V cB0 0683h 600EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3XU cB0 0683h 600E/100 256 N/A ECC B FC-PGA

(370 pin)

SL45U cB0 0683h 600E/100 256 N/A ECC B FC-PGA

(370 pin)

SL3XV cB0 0683h 650/100 256 N/A ECC B FC-PGA

(370 pin)

SL45W cB0 0683h 650/100 256 N/A ECC B FC-PGA

(370 pin)

SL3XW cB0 0683h 667/133 256 N/A ECC B FC-PGA

(370 pin)

SL45X cB0 0683h 667/133 256 N/A ECC B FC-PGA

(370 pin)

SL3XX cB0 0683h 700/100 256 N/A ECC B FC-PGA

(370 pin)

SL45Y cB0 0683h 700/100 256 N/A ECC B FC-PGA

(370 pin)

SL45Z cB0 0683h 733/133 256 N/A ECC B FC-PGA

(370 pin)

SL3XY cB0 0683h 733/133 256 N/A ECC B FC-PGA

(370 pin)

SL3XZ cB0 0683h 750/100 256 N/A ECC B FC-PGA

(370 pin)

SL462 cB0 0683h 750/100 256 N/A ECC B FC-PGA

(370 pin)

SL3Y2 cB0 0683h 800EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL464 cB0 0683h 800EB/133 256 N/A ECC B FC-PGA

(370 pin)

SL3Y3 cB0 0683h 800/100 256 N/A ECC B FC-PGA

(370 pin)

10, 7

7, 10

Specification Update 29

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL463 cB0 0683h 800/100 256 N/A ECC B FC-PGA

(370 pin)

SL43H cB0 0683h 850/100 256 N/A ECC B FC-PGA

(370 pin)

SL49G cB0 0683h 850/100 256 N/A ECC B FC-PGA

(370 pin)

SL43J cB0 0683h 866/133 256 N/A ECC B FC-PGA

(370 pin)

SL49H cB0 0683h 866/133 256 N/A ECC B FC-PGA

(370 pin)

SL44J cB0 0683h 933/133 256 N/A ECC B FC-PGA

(370 pin)

SL4CM cC0 0686h 600E/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CL cC0 0686h 600EB/133 256 N/A ECC C FC-PGA

(370 pin)

SL4CK cC0 0686h 650/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CJ cC0 0686h 667/133 256 N/A ECC C FC-PGA

(370 pin)

SL4CH cC0 0686h 700/100 256 N/A ECC C FC-PGA

(370 pin)

SL4M7 cC0 0686h 700/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CG cC0 0686h 733/133 256 N/A ECC C FC-PGA

(370 pin)

SL4M8 cC0 0686h 733/133 256 N/A ECC C FC-PGA

(370 pin)

SL4CF cC0 0686h 750/100 256 N/A ECC C FC-PGA

(370 pin)

SL4M9 cC0 0686h 750/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CE cC0 0686h 800/100 256 N/A ECC C FC-PGA

(370 pin)

SL4MA cC0 0686h 800/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CD cC0 0686h 800EB/133 256 N/A ECC C FC-PGA

(370 pin)

SL4MB cC0 0686h 800EB/133 256 N/A ECC C FC-PGA

(370 pin)

7, 10

7, 13

30 Specification Update

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL4CC cC0 0686h 850/100 256 N/A ECC C FC-PGA

(370 pin)

SL4MC cC0 0686h 850/100 256 N/A ECC C FC-PGA

(370 pin)

SL4CB cC0 0686h 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL4MD cC0 0686h 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL4SD cC0 0686h 900/100 256 N/A ECC C FC-PGA

(370 pin)

SL4C9 cC0 0686h 933/133 256 N/A ECC C FC-PGA

(370 pin)

SL4ME cC0 0686h 933/133 256 N/A ECC C FC-PGA

(370 pin)

SL4C8 cC0 0686h 1B GHz/133 256 N/A ECC C FC-PGA

(370 pin)

SL4MF cC0 0686h 1B GHz/133 256 N/A ECC C FC-PGA

(370 pin)

SL4WM cC0 0686h 1B GHz/133 256 N/A ECC C FC-PGA

(370 pin)

SL5BT cD0 068Ah 600E/100 256 N/A ECC C FC-PGA

(370 pin)

SL4ZM cD0 068Ah 700/100 256 N/A ECC C FC-PGA

(370 pin)

SL4ZL cD0 068Ah 733/133 256 N/A ECC C FC-PGA

(370 pin)

SL4Z4 cD0 068Ah 750/100 256 N/A ECC C FC-PGA

(370 pin)

SL4ZN cD0 068Ah 800/100 256 N/A ECC C FC-PGA

(370 pin)

SL52P cD0 068Ah 800EB/133 256 N/A ECC C FC-PGA

(370 pin)

SL5QD cD0 068Ah 800EB/133 256 N/A ECC C FC-PGA 2

(370 pin)

SL4Z2 cD0 068Ah 850/100 256 N/A ECC C FC-PGA

(370 pin)

SL49G cD0 068Ah 850/100 256 N/A ECC C FC-PGA

(370 pin)

SL5BS cD0 068Ah 900/100 256 N/A ECC C FC-PGA

(370 pin)

7, 13

7, 15

N/A

7, 17

Specification Update 31

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL4ZJ cD0 068Ah 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL49H cD0 068Ah 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL5B5/ cD0 068Ah 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL5DX cD0 068Ah 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL5QE cD0 068Ah 866/133 256 N/A ECC C FC-PGA

(370 pin)

SL52Q cD0 068Ah 933/133 256 N/A ECC C FC-PGA

(370 pin)

SL5DW cD0 068Ah 933/133 256 N/A ECC C FC-PGA

(370 pin)

SL5U3 cD0 068Ah 933MHz

/133

SL5QV cD0 068Ah 1GHz /100 256 N/A ECC C FC-PGA

SL52R cD0 068Ah 1BGHz/133 256 N/A ECC C FC-PGA

SL4F9 cD0 068Ah 1BGHz/133 256 N/A ECC C FC-PGA

SL5DV cD0 068Ah 1BGHz/133 256 N/A ECC C FC-PGA

SL5QW cD0 068Ah 1.10GHz

/100

SL5B3 cD0 068Ah 1BGHz/133 256 N/A ECC C FC-PGA 2

SL5FQ cD0 068Ah 1BGHz/133 256 N/A ECC C FC-PGA 2

SL5B5 cD0 068Ah 866/133 256 N/A ECC C FC-PGA 2

SL5QF cD0 068Ah 933/133 256 N/A ECC C FC-PGA

SL5QJ cD0 068Ah 1B GHz/133 256 N/A ECC C FC-PGA2

SL4YV cD0 068Ah 1.13

GHz/133

SL5B2 cD0 068Ah 1.13

GHz/133

256 N/A ECC C FC-PGA2

(370 pin)

256 N/A ECC C FC-PGA

(370 pin)

256 N/A ECC C FC-PGA2

(370 pin)

256 N/A ECC C FC-PGA2

(370 pin)

7, 17

17, 18

7, 17

7, 19

7, 17

17, 18

18, 19

32 Specification Update

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL4YV cD0 068Ah 1.13

GHz/133

SL5QK cD0 068Ah 1.13GHz/13

SL5GN tA1 06B1h 1.2GHz /133 256 N/A ECC FC-PGA2

SL5PM tA1 06B1h 1.2GHz /133 256 N/A ECC FC-PGA2

SL5GQ tA1 06B1h 1.13GHz/13

SL5LT tA1 06B1h 1.13GHz/13

SL5VX tA1 06B1h 1.333GHz

/133

SL64W tA1 06B1h

SL5GR tA1 06B1h 1GHz /133 256 N/A ECC FC-PGA2

SL66D tA1 06B1h 800 MHz-S

SL6HC tB1 06B4h 800 MHz-S

SL5PU tA1 06B1h 1.13GHz-S

SL5LV tA1 06B1h 1.13GHz-S

SL5QL tA1 06B1h 1.26 GHz-S

SL5LW tA1 06B1h 1.26 GHz-S

SL5LV tA1 06B1h 1.13GHz-S

SL657 tA1 06B1h 1.4 GHz-S

SL6BW tB1 06B4h 1.13GHz-S

SL6BX tB1 06B4h 1.26GHz-S

SL6BY tB1 06B4h 1.4GHz-S

1.40

GHz/133

/133

256 N/A ECC C FC-PGA

xxxiii(370

pin)

256 N/A ECC C FC-PGA2

(370 pin)

256 N/A ECC FC-PGA2

(370 pin)

256 N/A ECC FC-PGA2

(370 pin)

256 N/A ECC FC-PGA2

(370 pin)

256 No ECC FC-PGA2 9

(370 pin)

512 N/A ECC uFC-BGA 20, 21

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

512 N/A ECC FC-PGA2

(370 pin)

18, 19,20

7, 9

7, 20

Specification Update 33

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

SL3H7 cA2 0681h 600EB/133 256 N/A ECC B SECC2 10,20

SL3NB cA2 0681h 600EB/133 256 N/A ECC B SECC2 8,10,7,20

SL3KV cA2 0681h 650/100 256 N/A ECC B SECC2 10,7, 20

SL3NR cA2 0681h 650/100 256 N/A ECC B SECC2 8,10, 20

SL3KW cA2 0681h 667/133 256 N/A ECC B SECC2 7, 20

SL3SY cA2 0681h 700/100 256 N/A ECC B SECC2 8,10

SL3SB cA2 0681h 700/100 256 N/A ECC B SECC2 8,10

SL3S9 cA2 0681h 700/100 256 N/A ECC B SECC2 8,10

SL3SZ cA2 0681h 733/133 256 N/A ECC B SECC2 8,10

SL3H6 cA2 0681h 600E/100 256 N/A ECC B SECC2 8,10

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL3SB cA2 0681h 733/133 256 N/A ECC B SECC2 8,10

SL3ND cA2 0681h 667/133 256 N/A ECC B SECC2 8,10

SL3N6 cA2 0681h 533EB/133 256 N/A ECC B SECC2 8,10

SL3SX cA2 0681h 533EB/133 256 N/A ECC B SECC2 8,10

SL3V5 cA2 0681h 550E/100 256 N/A ECC B SECC2 8,10

SL3N7 cA2 0681h 550E/100 256 N/A ECC B SECC2 8,10

SL3NA cA2 0681h 600E/100 256 N/A ECC B SECC2 8,10

SL3WC cA2c 0681h 750/100 256 N/A ECC B SECC2 8,10

SL3V6 cA2 0681h 750/100 256 N/A ECC B SECC2 8,10

SL3Z6 cA2 0681h 800/100 256 N/A ECC B SECC2 8,10

SL3V7 cA2 0681h 800/100 256 N/A ECC B SECC2 8,10

SL3WA cA2 0681h 800EB/133 256 N/A ECC B SECC2 8,10

SL3V8 cA2 0681h 800EB/133 256 N/A ECC B SECC2 8,10

SL4G7 cA2 0681h 800EB/133 256 N/A ECC B SECC2 8,10

SL3XG cB0 0683h 533EB/133 256 N/A ECC B SECC2 8, 10

SL44W cB0c 0683h 533EB/133 256 N/A ECC B SECC2 8,10

SL3XH cB0 0683h 550E/100 256 N/A ECC B SECC2 8,10

SL44X cB0 0683h 550E/100 256 N/A ECC B SECC2 8,10

SL43E cB0 0683h 600E/100 256 N/A ECC B SECC2 8,10

SL44Y cB0 0683h 600E/100 256 N/A ECC B SECC2 8,10

34 Specification Update

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

SL3XJ cB0 0683h 600EB/133 256 N/A ECC B SECC2 8, 10

SL44Z cB0 0683h 600EB/133 256 N/A ECC B SECC2 8, 10

SL3XK cB0 0683h 650/100 256 N/A ECC B SECC2 8, 10

SL452 cB0 0683h 650/100 256 N/A ECC B SECC2 8, 10

SL3XL cB0 0683h 667/133 256 N/A ECC B SECC2 8, 10

SL453 cB0 0683h 667/133 256 N/A ECC B SECC2 8, 10

SL3XM cB0 0683h 700/100 256 N/A ECC B SECC2 8, 10

SL454 cB0 0683h 700/100 256 N/A ECC B SECC2 8, 10

SL3XN cB0 0683h 733/133 256 N/A ECC B SECC2 8, 10

SL455 cB0 0683h 733/133 256 N/A ECC B SECC2 8, 10

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL3XP cB0 0683h 750/100 256 N/A ECC B SECC2 8, 10

SL456 cB0 0683h 750/100 256 N/A ECC B SECC2 8, 10

SL3XQ cB0 0683h 800EB/133 256 N/A ECC B SECC2 8, 10

SL458 cB0 0683h 800EB/133 256 N/A ECC B SECC2 8, 10

SL3XR cB0 0683h 800/100 256 N/A ECC B SECC2 8, 10

SL457 cB0 0683h 800/100 256 N/A ECC B SECC2 8, 10

SL43F cB0 0683h 850/100 256 N/A ECC B SECC2 8, 10

SL47M cB0 0683h 850/100 256 N/A ECC B SECC2 8, 10

SL43G cB0 0683h 866/133 256 N/A ECC B SECC2 8, 10

SL47N cB0 0683h 866/133 256 N/A ECC B SECC2 8, 10

SL448 cB0 0683h 933/133 256 N/A ECC B SECC2 8, 10

SL47Q cB0 0683h 933/133 256 N/A ECC B SECC2 8, 10

SL4FP cB0 0683h 1B GHz/133 256 N/A ECC B SECC2 8, 10

SL48S cB0 0683h 1B GHz /133 256 N/A ECC B SECC2 8, 10

SL4C7 cC0 0686h 600E/100 256 N/A ECC B SECC2 8, 10

SL4C6 cC0 0686h 600EB/133 256 N/A ECC B SECC2 8, 10

SL4C5 cC0 0686h 650/100 256 N/A ECC B SECC2 8, 10

SL4C4 cC0 0686h 667/133 256 N/A ECC B SECC2 8, 10

SL4C3 cC0 0686h 700/100 256 N/A ECC B SECC2 8, 10

SL4C2 cC0 0686h 733/133 256 N/A ECC B SECC2 8, 10

Specification Update 35

Identification Information

Pentium® III Processor Identificati on and Package Information

S-Spec

SL4KD cC0 0686h 733/133 256 N/A ECC B SECC2 8, 10

SL4FQ cC0 0686h 733/133 256 N/A ECC B SECC2 8, 10

SL4BZ cC0 0686h 750/100 256 N/A ECC B SECC2 8, 10

SL4BY cC0 0686h 800/100 256 N/A ECC B SECC2 8, 10

SL4KF cC0 0686h 800/100 256 N/A ECC B SECC2 8, 10

SL4BX cC0 0686h 800EB/133 256 N/A ECC B SECC2 8, 10

SL4G7 cC0 0686h 800EB/133 256 N/A ECC B SECC2 8, 10

SL4KG cC0 0686h 800EB/133 256 N/A ECC B SECC2 8, 10

SL4BW cC0 0686h 850/100 256 N/A ECC B SECC2 8, 10

SL4KH cC0 0686h 850/100 256 N/A ECC B SECC2 8, 10

Core

Stepping CPUID

Speed (MHz)

Core/Bus11

L2 Size

(Kbytes)

Tag

RAM/

Stepping

ECC/ Non-

ECC

Processor Substrate

Revision

Package

and

Revision Notes

SL4BV cC0 0686h 866/133 256 N/A ECC B SECC2 8, 10

SL4KJ cC0 0686h 866/133 256 N/A ECC B SECC2 8, 10

SL4BT cC0 0686h 933/133 256 N/A ECC B SECC2 8, 10

SL4KK cC0 0686h 933/133 256 N/A ECC B SECC2 8, 10

SL4BR cC0 0686h 1 GHz/100 256 N/A ECC C SECC2 15

SL4KL cC0 0686h 1 GHz/100 256 N/A ECC C SECC2 8

SL4BS cC0 0686h 1B GHz/133 256 N/A ECC C SECC2 10,15

SL4HH cC0 0686h 1.13

GHz/133

SL69K tB1 06B4h 933 MHz-

S/133

SL6QU tB1 06B4h 1Ghz 512 N/A ECC N/A uFC-BGA 20,21

†

Unless otherwise noted, all Pentium III processors in S.E.C.C.2 package have an OLGA package core.

NOTES:

1. These parts will only operate at the specified core to bus frequency ratio at which they

were manufactured and tested. It is not necessary to configure the core frequency ratios by using the A20M#, IGNEE#, LINT[1]/NMI and LINT[0]/INTR pins during RESET.

2. These processors will not shut down automatically upon assertion of THERMTRIP#.

3. This is a boxed processor with an attached heatsink.

4. Performance-monitoring event counters do not reflect MOVD and MOVQ stores to

memory on these processors.

5. These parts will not assert THERMTRIP#, nor will they shut down in the event of an

over-temperature condition (e.g., Tj = ~135 C).

6. Pin AJ3 is removed from these parts.

7. This is a boxed processor with an unattached fan heatsink.

8. This is a boxed processor with an attached fan heatsink.

9. These processors will not be validated in Dual Processor (DP) applications.

256 N/A ECC C SECC2 12,8

512 N/A ECC N/A uFC-BGA 20,21

36 Specification Update

Identification Information

10. The “E” and “B” designators distinguish between Pentium® III processors with the

same core frequency but different system bus frequencies and/or cache implementations. The “E” and “B” designators distinguish between Pentium® III processors with the same core frequency but different system bus frequencies and/or cache implementations. B = 133 MHz System Bus E = Processors with “Advanced Transfer Cache” (CPUID 068x and greater only if a frequency overlap exists) If, for a given core frequency, Pentium III processors are only available with one system bus frequency and one cache implementation, the above designators will not be used (e.g., not all processors with “Advanced Transfer Cache” will have the “E” designation).

11. Speeds will be marked as MHz up to but not including 1GHz. Speeds 1GHz and above

will have the GHz marking.

12. Vcc = 1.80V. Tj = 60C for this 1.13 GHz processor with CPUID 0686.

13. Tj = 80C, Vcc = 1.70v.

14. Vcc = 1.65V.

15. Vcc = 1.70V for these cCX core steppings. Tj = 70C for 1.0 GHz. Tj = 75C for 933

MHz.

16. This SL4WM S-spec part has a VID request of 1.70V, however the processor should be

supplie d 1.76V at the PGA V cc pins. See Pentium® III datasheet for further information.

17. Vcc=1.75V for cD0 Core Stepping (CPUID 068Ah). Tj=77 C for 933MHz and Tj=75 C

1GHz. Tj=80 C for 866MHz to 700MHz.

18. This processor is valid for low voltage system bus operation at 1.25V AGTL and normal

1.5V AGTL+ signal levels. This processor is also DP capable at the 1.25V AGTL system bus level. This processor will auto detect differential or single ended clocking.

19. Vcc=1.75V for cD0 Core Stepping (CPUID 068Ah). Tcase=64 C for 1GHz. Tcase = 67

C for 1.13GHz. This package exists as an FC-PGA2 with Integrated Heat Spreader (IHS).

20. These parts are intended for server design applications.

21. Tualatin LV DP 1.15V, non SpeedStep enabled

Component Marking Information

Pentium® III Processor and Boxed Pentium® III Processor Markings

Dynamic Mark Area

FPO - Serial #

of Assy

Specification Update 37

Country

Speed / Cache / Bus / Voltage

2-D Matrix Mark

UL Identifier

500/ 512/100/2.0V S1 FFFFFFFF-NNNN XXXXX

i ©’98 SYYYY

S-Spec

Pentium® III Processor Markings

SECC2/Slot 1 Package

FC-PGA 370 Pin Package

Identification Information

Hologram Location

GRP1LN1: INTEL (m)(c) '01_-_{COO} GRP1LN2: {Speed}/{Cache}/{Bus}/{Voltage}

FC-PGA2 370 Pin Package

GRP1LN1 GRP1LN2

GRP2LN1 GRP2LN2

GRP2LN1: {FPO}-{S/N} GRP2LN2: PENTIUM III {S-Spec}

GRP1LN1: INTEL (m)(c) '01_-_{Country of Origin} GRP1LN2: {Core freq}/{Cache}/{Bus Freq}/{Voltage}

GRP2LN1: {FPO}-{S/N}

GRP2LN2: PENTIUM III {S-Spec} or PENTIUM III-S {S-Spec}

Note: S above applies to 06BxH 512KB cache processor

38 Specification Update

Errata

E1. FP Data Operand Pointer May Be Incorrectly Calculated After FP

Access Which Wraps 64 Kbyt e Bound ary in 16 Bit Code

Problem: The FP Data Operand Pointer is the effective address of the operand associated with

the last non-control floating-point instruction executed by the machine. If an 80-bit floating-point access (load or store) occurs in a 16-bit mode other than protected mode (in which case the access will produce a segment limit violation), the memory access wraps a 64-Kbyte boundary, and the floating-point environment is subsequently saved, the value contained in the FP Data Operand Pointer may be incorrect.

Implication: A 32-bit operating system running 16-bit floating-point code may encounter this

erratum, under the following conditions:

• The operating system is using a segment greater than 64 Kbytes in size.

• An application is running in a 16-bit mode other than protected mode.

• An 80-bit floating-point load or store which wraps the 64-Kbyte boundary is executed.

• The operating system performs a floating-point environment store (FSAVE/FNSAVE/FSTENV/FNSTENV) after the above memory access.

• The operating system uses the value contained in the FP Data Operand Pointer.

Wrapping an 80 bit floating-point load around a segment boundary in this way is not a normal programming practice. Intel has not currently identified any software which exhibits this behavior.

Workaround: If the FP Data Operand Pointer is used in an OS which may run 16-bit floating-

point code, care must be taken to ensure that no 80-bit floating-point accesses are wrapped around a 64-Kbyte boundary

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E2. Differences Exist in Debug Exception Reporting

Problem: There exist some differences in the reporting of code and data breakpoint matches

between that specified by previous Intel processor specifications and the behavior of the processor, as described below:

Case 1: The first case is for a breakpoint set on a MOVSS or POPSS instruction, when

the instruction following it causes a debug register protection fault (DR7.gd is already set, enabling the fault). The processor reports delayed data breakpoint matches from the MOVSS or POPSS instructions by setting the matching DR6.bi bits, along with the debug register protection fault

Specification Update 39

Errata

(DR6.bd). If additional breakpoint faults are matched during the call of the debug fault handler, the processor sets the breakpoint match bits (DR6.bi) to reflect the breakpoints matched by both the MOVSS or POPSS breakpoint and the debug fault handler call. The processor only sets DR6.bd in either situation, and does not set any of the DR6.bi bits.

Case 2: In the second breakpoint reporting failure case, if a MOVSS or POPSS

instruction with a data breakpoint is followed by a store to memory which:

a) crosses a 4-Kbyte page boundary,

b) causes the page table Access or Dirty (A/D) bits to be modified,

the breakpoint information for the MOVSS or POPSS will be lost. Previous processors retain this information under these boundary conditions.

Case 3: If they occur after a MOVSS or POPSS instruction, the INTn, INTO, and INT3

instructions zero the DR6.bi bits (bits B0 through B3), clearing pending breakpoint information, unlike previous processors.

Case 4: If a data breakpoint and an SMI (System Management Interrupt) occur

simultaneously, the SMI will be serviced via a call to the SMM handler, and the pending breakpoint will be lost.

Case 5: When an instruction that accesses a debug register is executed, and a

breakpoint is encountered on the instruction, the breakpoint is reported twice.

Case 6: Unlike previous versions of Intel Architecture processors, P6 family

processors will not set the Bi bits for a matching disabled breakpoint unless at least one other breakpoint is enabled.

Implication: When debugging or when developing debuggers for a P6 family processor-based

system, this behavior should be noted. Normal usage of the MOVSS or POPSS instructions (i.e., following them with a MOV ESP) will not exhibit the behavior of cases 1-3. Debugging in conjunction with SMM will be limited by case 4.

Workaround: Following MOVSS and POPSS instructions with a MOV ESP instruction when using

breakpoints will avoid the first three cases of this erratum. No workaround has been identified for cases 4, 5, or 6.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E3. FLUSH# S ervicing Delayed While Waiting for STA RTUP_IPI in 2 way

MP Systems

Problem: In a 2-way MP system, if an application processor is waiting for a startup inter-

processor interrupt (STARTUP_IPI), then it will not service a FLUSH# pin assertion until it has received the STARTUP_IPI.

Implication: After the 2-way MP initialization protocol, only one processor becomes the bootstrap

processor (BSP). The other processor becomes a slave application processor (AP).

40 Specification Update

Errata

After losing the BSP arbitration, the AP goes into a wait loop, waiting for a STARTUP_IPI.

The BSP can wake up the AP to perform some tasks with a STARTUP_IPI, and then put it back to sleep with an initialization inter-processor interrupt (INIT_IPI, which has the same effect as asserting INIT#), which returns it to a wait loop. The result is a possible loss of cache coherency if the off-line processor is intended to service a FLUSH# assertion at this point. The FLUSH# will be serviced as soon as the processor is awakened by a STARTUP_IPI, before any other instructions are executed. Intel has not encountered any operating systems that are affected by this erratum.

Workaround: Operating system developers should take care to execute a WBINVD instruction

before the AP is taken off-line using an INIT_IPI

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E4. Code Fetch Matching Disabled Debug Register May Cause Debug

Exception

Problem: The bits L0 3 and G0 3 enable breakpoints local to a task and global to all tasks,

respectively. If one of these bits is set, a breakpoint is enabled, corresponding to the addresses in the debug registers DR0-DR3. If at least one of these breakpoints is enabled, any of these registers are disabled (i.e., Ln and Gn are 0), and RWn for the disabled register is 00 (indicating a breakpoint on instruction execution), normally an instruction fetch will not cause an instruction-breakpoint fault based on a match with the address in the disabled register(s). However, if the address in a disabled register matches the address of a code fetch which also results in a page fault, an instructionbreakpoint fault will occur.

Implication: The bits L0 3 and G0 3 enable breakpoints local to a task and global to all tasks,

Implication: While debugging software, extraneous instruction-breakpoint faults may be

encountered if breakpoint registers are not cleared when they are disabled. Debug software which does not implement a code breakpoint handler will fail, if this occurs. If a handler is present, the fault will be serviced. Mixing data and code may exacerbate this problem by allowing disabled data breakpoint registers to break on an instruction fetch.

Workaround: The debug handler should clear breakpoint registers before they become disabled Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E5. Double ECC Error on Read May Result in BINIT#

Specification Update 41

Errata

Problem: For this erratum to occur, the following conditions must be met:

• Machine Check Exceptions (MCEs) must be enabled.

• A dataless transaction (such as a write invalidate) must be occurring simultaneously with a transaction which returns data (a normal read).

• The read data must contain a double-bit uncorrectable ECC error.

If these conditions are met, the Pentium III processor will not be able to determine which transaction was erroneous, and instead of generating an MCE, it will generate a BINIT#.

Implication: The bus will be reinitialized in this case. However, since a double-bit uncorrectable

ECC error occurred on the read, the MCE handler (which is normally reached on a double-bit uncorrectable ECC error for a read) would most likely cause the same BINIT# event.

Workaround: Though the ability to drive BINIT# can be disabled in the Pentium III processor,

which would prevent the effects of this erratum, overall system behavior would not improve, since the error which would normally cause a BINIT# would instead cause the machine to shut down. No other workaround has been identified.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E6. FP Inexact-Result Exception Flag May Not Be Set

Problem: When the result of a floating-point operation is not exactly representable in the

destination format (1/3 in binary form, for example), an inexact-result (precision) exception occurs. When this occurs, the PE bit (bit 5 of the FPU status word) is normally set by the processor. Under certain rare conditions, this bit may not be set when this rounding occurs. However, other actions taken by the processor (invoking the software exception handler if the exception is unmasked) are not affected. This erratum can only occur if the floating-point operation which causes the precision exception is immediately followed by one of the following instructions:

• FST m32real

• FST m64real

• FSTP m32real

• FSTP m64real

• FSTP m80real

• FIST m16int

• FIST m32int

• FISTP m16int

• FISTP m32int

• FISTP m64int

42 Specification Update

Errata

Note that even if this combination of instructions is encountered, there is also a dependency on the internal pipelining and execution state of both instructions in the processor.

Implication: Inexact-result exceptions are commonly masked or ignored by applications, as it

happens frequently, and produces a rounded result acceptable to most applications. The PE bit of the FPU status word may not always be set upon receiving an inexactresult exception. Thus, if these exceptions are unmasked, a floating-point error exception handler may not recognize that a precision exception occurred. Note that this is a “sticky” bit, i.e., once set by an inexact-result condition, it remains set until cleared by software.

Workaround: This condition can be avoided by inserting two NOP instructions between the two

floating-point instructions.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E7. BTM for SMI Will Contain Incorrect FROM EIP

Problem: A system management interrupt (SMI) will produce a Branch Trace Message (BTM), if

BTMs are enabled. However, the FROM EIP field of the BTM (used to determine the address of the instruction which was being executed when the SMI was serviced) will not have been updated for the SMI, so the field will report the same FROM EIP as the previous BTM.

Implication: A BTM which is issued for an SMI will not contain the correct FROM EIP, limiting the

usefulness of BTMs for debugging software in conjunction with System Management Mode (SMM).

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E8. I/O Restart in SMM May Fail After Simultaneous MCE

Problem: If an I/O instruction (IN, INS, REP INS, OUT, OUTS, or REP OUTS) is being executed,

and if the data for this instruction becomes corrupted, the Pentium III processor will signal a machine check exception (MCE). If the instruction is directed at a device which is powered down, the processor may also receive an assertion of SMI#. Since MCEs have higher priority, the processor will call the MCE handler, and the SMI# assertion will remain pending. However, upon attempting to execute the first instruction of the MCE handler, the SMI# will be recognized and the processor will attempt to execute the SMM handler. If the SMM handler is completed successfully, it will attempt to restart the I/O instruction, but will not have the correct machine state, due to the call to the MCE handler.

Implication: A simultaneous MCE and SMI# assertion may occur for one of the I/O instructions

above. The SMM handler may attempt to restart such an I/O instruction, but will have corrupted state due to the MCE handler call, leading to failure of the restart and shutdown of the processor.

Specification Update 43

Errata

Workaround: If a system implementation must support both SMM and MCEs, the first thing the

SMM handler code (when an I/O restart is to be performed) should do is check for a pending MCE. If there is an MCE pending, the SMM handler should immediately exit via an RSM instruction and allow the machine check exception handler to execute. If there is not, the SMM handler may proceed with its normal operation.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E9. Branch Traps Do Not Function If BTMs Are Also Enabled

Problem: If branch traps or branch trace messages (BTMs) are enabled alone, both function as

expected. However, if both are enabled, only the BTMs will function, and the branch traps will be ignored.

Implication: The branch traps and branch trace message debugging features cannot be used

together.

Workaround: If branch trap functionality is desired, BTMs must be disabled. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E10. Checker BIST Failure in FRC Mode Not Signaled

Problem: If a system is running in functional redundancy checking (FRC) mode, and the

checker of the master-checker pair encounters a hard failure while running the builtin self test (BIST), the checker will tri-state all outputs without signaling an IERR#.

Implication: Assuming the master passes BIST successfully, it will continue execution unchecked,

operating without functional redundancy. However, the necessary pull-up on the FRCERR pin will cause an FRCERR to be signaled. The operation of the master depends on the implementation of FRCERR.

Workaround: For successful detection of BIST failure in the checker of an FRC pair, use the

FRCERR signal, instead of IERR#.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E11. BINIT# Assertion Causes FRCERR Assertion in FRC Mode

Problem: If a pair of Pentium III processors are running in functional redundancy checking

(FRC) mode, and a catastrophic error condition causes BINIT# to be asserted, the checker in the master-checker pair will enter shutdown. The next bus transaction from the master will then result in the assertion of FRCERR.

Implication: Bus initialization via an assertion of BINIT# occurs as the result of a catastrophic

error condition which precludes the continuing reliable execution of the system. Under normal circumstances, the master-checker pair would remain synchronized in the execution of the BINIT# handler. However, due to this erratum, an FRCERR will be signaled. System behavior then depends on the system specific error recovery mechanisms.

44 Specification Update

Errata

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E12. Machine Check Exception Handler May Not Always Execute

Successfully

Problem: An asynchronous machine check exception (MCE), such as a BINIT# event, which

occurs during an access that splits a 4 Kbyte page boundary, may leave some internal registers in an indeterminate state. Thus, the MCE handler code may not always run successfully if an asynchronous MCE has occurred previously.

Implication: An MCE may not always result in the successful execution of the MCE handler.

However, asynchronous MCEs usually occur upon detection of a catastrophic system condition that would also hang the processor. Leaving MCEs disabled will result in the condition which caused the asynchronous MCE instead causing the processor to enter shutdown. Therefore, leaving MCEs disabled may not improve overall system behavior.

Workaround: No workaround which would guarantee successful MCE handler execution under

this condition has been identified.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E13. MCE Due to L2 Parity Error Gives L1 MCACOD.LL

Problem: If a Cache Reply Parity (CRP) error, Cache Address Parity (CAP) error, or Cache

Synchronous Error (CSER) occurs on an access to the Pentium III processor’s L2 cache, the resulting Machine Check Architectural Error Code (MCACOD) will be logged with ‘01’ in the LL field. This value indicates an L1 cache error; the value should be ‘10’, indicating an L2 cache error. Note that L2 ECC errors have the correct value of ‘10’ logged.

Implication: An L2 cache access error, other than an ECC error, will be improperly logged as an L1

cache error in MCACOD.LL.

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E14. LBER May Be Corrupted After Some Events

Problem: The last branch record (LBR) and the last branch before exception record (LBER) can

be used to determine the source and destination information for previous branches or exceptions. The LBR contains the source and destination addresses for the last branch or exception, and the LBER contains similar information for the last branch taken before the last exception. This information is typically used to determine the location of a branch which leads to execution of code which causes an exception. However, after a catastrophic bus condition which results in an assertion of BINIT# and the re-initialization of the buses, the value in the LBER may be corrupted. Also,

Specification Update 45

Errata

after either a CALL which results in a fault or a software interrupt, the LBER and LBR will be updated to the same value, when the LBER should not have been updated.

Implication: The LBER and LBR registers are used only for debugging purposes. When this

erratum occurs, the LBER will not contain reliable address information. The value of LBER should be used with caution when debugging branching code; if the values in the LBR and LBER are the same, then the LBER value is incorrect. Also, the value in the LBER should not be relied upon after a BINIT# event.

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E15. BTMs May Be Corrupted During Simultaneous L1 Cache Line

Replacement

Problem: When Branch Trace Messages (BTMs) are enabled and such a message is generated,

the BTM may be corrupted when issued to the bus by the L1 cache if a new line of data is brought into the L1 data cache simultaneously. Though the new line being stored in the L1 cache is stored correctly, and no corruption occurs in the data, the information in the BTM may be incorrect due to the internal collision of the data line and the BTM.

Implication: Although BTMs may not be entirely reliable due to this erratum, the conditions

necessary for this boundary condition to occur have only been exhibited during focused simulation testing. Intel has currently not observed this erratum in a system level validation environment.

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E16. EFLAGS Discrepancy on a Page Fault After a Multiprocessor TLB

Shootdown

Problem: This erratum may occur when the Pentium III processor executes one of the following

read-modify-write arithmetic instructions and a page fault occurs during the store of the memory operand: ADD, AND, BTC, BTR, BTS, CMPXCHG, DEC, INC, NEG, NOT, OR, ROL/ROR, SAL/SAR/SHL/SHR, SHLD, SHRD, SUB, XOR, and XADD. In this case, the EFLAGS value pushed onto the stack of the page fault handler may reflect the status of the register after the instruction would have completed execution rather than before it. The following conditions are required for the store to generate a page fault and call the operating system page fault handler:

1. The store address entry must be evicted from the DTLB by speculative loads from other instructions that hit the same way of the DTLB before the store has completed. DTLB eviction requires at least three-load operations that have linear address bits 15:12 equal to each other and address bits 31:16 different from each other in close physical proximity to the arithmetic operation.

2. The page table entry for the store address must have its permissions tightened during the very small window of time between the DTLB eviction and execution of

46 Specification Update

Errata

the store. Examples of page permission tightening include from Present to Not Present or from Read/Write to Read Only, etc.

3. Another processor, without corresponding synchronization and TLB flush, must cause the permission change.

Implication: This scenario may only occur on a multiprocessor platform running an operating

system that performs “lazy” TLB shootdowns. The memory image of the EFLAGS register on the page fault handler’s stack prematurely contains the final arithmetic flag values although the instruction has not yet completed. Intel has not identified any operating systems that inspect the arithmetic portion of the EFLAGS register during a page fault nor observed this erratum in laboratory testing of software applications.

Workaround: No workaround is needed upon normal restart of the instruction, since this

erratum is transparent to the faulting code and results in correct instruction behavior. Operating systems may ensure that no processor is currently accessing a page that is scheduled to have its page permissions tightened or have a page fault handler that ignores any incorrect state.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E17. Near CALL to ESP Creates Unexpected EIP Address

Problem: As documented, the CALL instruction saves procedure linking information in the

procedure stack and jumps to the called procedure specified with the destination (target) operand. The target operand specifies the address of the first instruction in the called procedure. This operand can be an immediate value, a general-purpose register, or a memory location. When accessing an absolute address indirectly using the stack pointer (ESP) as a base register, the base value used is the value in the ESP register before the instruction executes. However, when accessing an absolute address directly using ESP as the base register, the base value used is the value of ESP after the return value is pushed on the stack, not the value in the ESP register before the instruction executed.

Implication: Due to this erratum, the processor may transfer control to an unintended address.

Results are unpredictable, depending on the particular application, and can range from no effect to the unexpected termination of the application due to an exception. Intel has observed this erratum only in a focused testing environment. Intel has not observed any commercially available operating system, application, or compiler that makes use of or generates this instruction.

Workaround: If the other seven general-purpose registers are unavailable for use, and it is

necessary to do a CALL via the ESP register, first push ESP onto the stack, then perform an indirect call using ESP (e.g., CALL [ESP]). The saved version of ESP should be popped off the stack after the call returns.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E18. Memory Type Undefined for Nonmemory Operations

Specification Update 47

Errata

Problem: The Memory Type field for nonmemory transactions such as I/O and Special Cycles

are undefined. Although the Memory Type attribute for nonmemory operations logically should (and usually does) manifest itself as UC, this feature is not designed into the implementation and is therefore inconsistent.

Implication: Bus agents may decode a non-UC memory type for nonmemory bus transactions. Workaround: Bus agents must consider transaction type to determine the validity of the

Memory Type field for a transaction.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E19. Infinite Snoop Stall During L2 Initialization of MP Systems

Problem: It is possible for snoop traffic generated on the system bus while a processor is

executing its L2 cache initialization routine to cause the initializing processor to hang.

Implication: A DP (2-way) system which does not suppress snoop traffic while L2 caches are being

initialized may hang during this initialization sequence.

The system BIOS can create an execution environment which allows processors to initialize their L2 caches without the system generating any snoop traffic on the bus.

Below is a pseudo-code fragment, designed explicitly for a two-processor system, that uses a serial algorithm to initialize each processor’s L2 cache:

Suppress_all_I/O_traffic()

K = 0;

while (K <= 1)

{

/* Obtain current value of K. This forces both Temp and K into */

/* the L1 cache. Note that Temp could also be maintained in a */

/* general purpose register. */

Temp = K;

Wait_until_all_processors_are_signed_in_at_barrier()

if ( logical_proc_APIC_id == K ) {

{

wait_10_usecs_delay_loop(); /* this time delay, required */

/* in the worst case, allows */

/* the barrier semaphore to */

48 Specification Update

Errata

/* settle to shared state. */

Initialize L2 cache

K++

}

else

while (Temp == K);

}

This algorithm prevents bus snoop traffic from the other processors, which would otherwise cause the initializing processor to hang. The algorithm assumes that the L1 cache is enabled (the Temp and K variables must be cached by each processor). Also, the Memory Type Range Register (MTRR) for the data segment must be set to WB (writeback) memory type.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E20. FP Data Operand Pointer May Not Be Zero After Power On or Reset

Problem: The FP Data Operand Pointer, as specified, should be reset to zero upon power on or

Reset by the processor. Due to this erratum, the FP Data Operand Pointer may be nonzero after power on or Reset.

Implication: Software which uses the FP Data Operand Pointer and count on its value being zero

after power on or Reset without first executing an FINIT/FNINIT instruction will use an incorrect value, resulting in incorrect behavior of the software.

Workaround: Software should follow the recommendation in Section 8.2 of the Intel

Architecture Software Developer’s Manual, Volume 3: System Programming Guide (Order Number 243192). This recommendation states that if the FPU will be used, software-initialization code should execute an FINIT/FNINIT instruction following a hardware reset. This will correctly clear the FP Data Operand Pointer to zero.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E21. MOVD Following Zeroing Instruction Can Cause Incorrect Result

Problem: An incorrect result may be calculated after the following circumstances occur:

1. A register has been zeroed with either a SUB reg, reg instruction or an XOR reg, reg instruction,

2. A value is moved with sign extension into the same register’s lower 16 bits; or a signed integer multiply is performed to the same register’s lower 16 bits,

3. This register is then copied to an MMX™ technology register using the MOVD instruction prior to any other operations on the sign-extended value.

Specification Update 49

Errata

Specifically, the sign may be incorrectly extended into bits 16-31 of the MMX technology register. Only the MMX technology register is affected by this erratum.

The erratum only occurs when the three following steps occur in the order shown. The erratum may occur with up to 40 intervening instructions that do not modify the sign-extended value between steps 2 and 3.

1. XOR EAX, EAX

or SUB EAX, EAX

2. MOVSX AX, BL

or MOVSX AX, byte ptr <memory address> or MOVSX AX, BX

or MOVSX AX, word ptr <memory address> or IMUL BL (AX implicit, opcode F6 /5)

or IMUL byte ptr <memory address> (AX implicit, opcode F6 /5) or IMUL AX, BX (opcode 0F AF /r)

or IMUL AX, word ptr <memory address> (opcode 0F AF /r) or IMUL AX, BX, 16 (opcode 6B /r ib)

or IMUL AX, word ptr <memory address>, 16 (opcode 6B /r ib) or IMUL AX, 8 (opcode 6B /r ib)

or IMUL AX, BX, 1024 (opcode 69 /r iw)

or IMUL AX, word ptr <memory address>, 1024 (opcode 69 /r iw) or IMUL AX, 1024 (opcode 69 /r iw)

or CBW

3. MOVD MM0, EAX

Note that the values for immediate byte/words are merely representative (i.e., 8, 16,

1024) and that any value in the range for the size may be affected. Also, note that

this erratum may occur with “EAX” replaced with any 32 bit general-purpose register, and “AX” with the corresponding 16-bit version of that replacement. “BL” or “BX” can be replaced with any 8-bit or 16-bit general-purpose register. The CBW and IMUL (opcode F6 /5) instructions are specific to the EAX register only.

In the example, EAX is forced to contain 0 by the XOR or SUB instructions. Since the four types of the MOVSX or IMUL instructions and the CBW instruction modify only bits 15:8 of EAX by sign extending the lower 8 bits of EAX, bits 31:16 of EAX should always contain 0. This implies that when MOVD copies EAX to MM0, bits 31:16 of MM0 should also be 0. Under certain scenarios, bits 31:16 of MM0 are not 0, but are replicas of bit 15 (the 16th bit) of AX. This is noticeable when the value in AX after the MOVSX, IMUL, or CBW instruction is negative, i.e., bit 15 of AX is a 1.

When AX is positive (bit 15 of AX is a 0), MOVD will always produce the correct answer. If AX is negative (bit 15 of AX is a 1), MOVD may produce the right answer or the wrong answer depending on the point in time when the MOVD instruction is executed in relation to the MOVSX, IMUL, or CBW instruction.

50 Specification Update

Errata

Implication: The effect of incorrect execution will vary from unnoticeable, due to the code

sequence discarding the incorrect bits, to an application failure. If the MMX technology-enabled application in which MOVD is used to manipulate pixels, it is possible for one or more pixels to exhibit the wrong color or position momentarily. It is also possible for a computational application that uses the MOVD instruction in the manner described above to produce incorrect data. Note that this data may cause an unexpected page fault or general protection fault

Workaround: There are two possible workarounds for this erratum:

1. Rather than using the MOVSX-MOVD, IMUL-MOVD, or CBW-MOVD pairing to handle one variable at a time, use the sign extension capabilities (PSRAW, etc.) within MMX technology for operating on multiple variables. This would result in higher performance as well.

2. Insert another operation that modifies or copies the sign-extended value between the MOVSX/IMUL/CBW instruction and the MOVD instruction as in the example below:

XOR EAX, EAX (or SUB EAX, EAX)

MOVSX AX, BL (or other MOVSX, other IMUL or CBW instruction)

*MOV EAX, EAX

MOVD MM0, EAX

*Note: MOV EAX, EAX is used here as it is fairly generic. Again, EAX can be any 32bit register.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E22. Premature Execution of a Load Operation Prior to Exception Handler

Invocation

Problem: This erratum can occur with any of the following situations:

1. If an instruction that performs a memory load causes a code segment limit violation,

2. If a waiting floating-point instruction or MMX™ instruction that performs a memory load has a floating-point exception pending, or

3. If an MMX instruction that performs a memory load and has either CR0.EM =1 (Emulation bit set), or a floating-point Top-of-Stack (FP TOS) not equal to 0, or a DNA exception pending.

If any of the above circumstances occur it is possible that the load portion of the instruction will have executed before the exception handler is entered.

Implication: In normal code execution where the target of the load operation is to write back

memory there is no impact from the load being prematurely executed, nor from the restart and subsequent re-execution of that instruction by the exception handler. If the target of the load is to uncached memory that has a system side-effect,

Specification Update 51

Errata

restarting the instruction may cause unexpected system behavior due to the repetition of the side-effect.

Workaround: Code which performs loads from memory that has side-effects can effectively

workaround this behavior by using simple integer-based load instructions when accessing side-effect memory and by ensuring that all code is written such that a code segment limit violation cannot occur as a part of reading from side-effect memory.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E23. Read Portion of RMW Instruction May Execute Twice

Problem: When the Pentium III processor executes a read-modify-write (RMW) arithmetic

instruction, with memory as the destination, it is possible for a page fault to occur during the execution of the store on the memory operand after the read operation has completed but before the write operation completes.

If the memory targeted for the instruction is UC (uncached), memory will observe the occurrence of the initial load before the page fault handler and again if the instruction is restarted.

Implication: This erratum has no effect if the memory targeted for the RMW instruction has no

side-effects. If, however, the load targets a memory region that has side-effects, multiple occurrences of the initial load may lead to unpredictable system behavior.

Workaround: Hardware and software developers who write device drivers for custom hardware

that may have a side-effect style of design should use simple loads and simple stores to transfer data to and from the device. Then, the memory location will simply be read twice with no additional implications.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E24. MC2_STATUS MSR Has Model-Specific Error Code and Machine Check

Architecture Error Code Reversed

Problem: The Intel Architecture Software Developer’s Manual, Volume 3: System Programming

Guide, documents that for the MCi_STATUS MSR, bits 15:0 contain the MCA (machine-check architecture) error code field, and bits 31:16 contain the modelspecific error code field. However, for the MC2_STATUS MSR, these bits have been reversed. For the MC2_STATUS MSR, bits 15:0 contain the model-specific error code field and bits 31:16 contain the MCA error code field.

Implication: A machine check error may be decoded incorrectly if this erratum on the

MC2_STATUS MSR is not taken into account.

Workaround: When decoding the MC2_STATUS MSR, reverse the two error fields. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E25. Mixed Cacheability of Lock Variables Is Problematic in MP Systems

52 Specification Update

Errata

Problem: This errata only affects multiprocessor systems where a lock variable address is

marked cacheable in one processor and uncacheable in any others. The processors which have it marked uncacheable may stall indefinitely when accessing the lock variable. The stall is only encountered if:

• One processor has the lock variable cached, and is attempting to execute a cache lock.

• If the processor which has that address cached has it cached in its L2 only.

• Other processors, meanwhile, issue back to back accesses to that same address on the bus.

Implication: MP systems where all processors either use cache locks or consistent locks to

uncacheable space will not encounter this problem. If, however, a lock variable’s cacheability varies in different processors, and several processors are all attempting to perform the lock simultaneously, an indefinite stall may be experienced by the processors which have it marked uncacheable in locking the variable (if the conditions above are satisfied). Intel has only encountered this problem in focus testing with artificially generated external events. Intel has not currently identified any commercial software which exhibits this problem.

Workaround: Follow a homogenous model for the memory type range registers (MTRRs),

ensuring that all processors have the same cacheability attributes for each region of memory; do not use locks whose memory type is cacheable on one processor, and uncacheable on others. Avoid page table aliasing, which may produce a nonhomogenous memory model.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E26. MOV With Debug Register Causes Debug Exception

Problem: When in V86 mode, if a MOV instruction is executed on debug registers, a general-

protection exception (#GP) should be generated, as documented in the Intel Architecture Software Developer's Manual, Volume 3: System Programming Guide, Section 14.2. However, in the case when the general detect enable flag (GD) bit is set, the observed behavior is that a debug exception (#DB) is generated instead.

Implication: With debug-register protection enabled (i.e., the GD bit set), when attempting to

execute a MOV on debug registers in V86 mode, a debug exception will be generated instead of the expected general-protection fault.

Workaround: In general, operating systems do not set the GD bit when they are in V86 mode.

The GD bit is generally set and used by debuggers. The debug exception handler should check that the exception did not occur in V86 mode before continuing. If the exception did occur in V86 mode, the exception may be directed to the generalprotection exception handler.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E27. Upper Four PAT Entries Not Usable With Mode B or Mode C Paging

Specification Update 53

Errata

Problem: The Page Attribute Table (PAT) contains eight entries, which must all be initialized

and considered when setting up memory types for the Pentium III processor. However, in Mode B or Mode C paging, the upper four entries do not function correctly for 4-Kbyte pages. Specifically, bit 7 of page table entries that translate addresses to 4-Kbyte pages should be used as the upper bit of a 3-bit index to determine the PAT entry that specifies the memory type for the page. When Mode B (CR4.PSE = 1) and/or Mode C (CR4.PAE) are enabled, the processor forces this bit to zero when determining the memory type regardless of the value in the page table entry. The upper four entries of the PAT function correctly for 2-Mbyte and 4-Mbyte large pages (specified by bit 12 of the page directory entry for those translations).

Implication: Only the lower four PAT entries are useful for 4-KB translations when Mode B or C

paging is used. In Mode A paging (4-Kbyte pages only), all eight entries may be used. All eight entries may be used for large pages in Mode B or C paging.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E28. Dat a Breakpoint Exce ption in a Displace ment Relative Near Call May

Corrupt EIP

Problem: If a misaligned data breakpoint is programmed to the same cache line as the

memory location where the stack push of a near call is performed and any data breakpoints are enabled, the processor will update the stack and ESP appropriately, but may skip the code at the destination of the call. Hence, program execution will continue with the next instruction immediately following the call, instead of the target of the call.

Implication: The failure mechanism for this erratum is that the call would not be taken; therefore,

instructions in the called subroutine would not be executed. As a result, any code relying on the execution of the subroutine will behave unpredictably.

Workaround: Whether enabled or not, do not program a misaligned data breakpoint to the

same cache line on the stack where the push for the near call is performed.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E29. RDMSR or WRMSR to Invalid MSR Address May Not Cause GP F a ul t

Problem: The RDMSR and WRMSR instructions allow reading or writing of MSRs (Model Specific

Registers) based on the index number placed in ECX. The processor should reject access to any reserved or unimplemented MSRs by generating #GP(0). However, there are some invalid MSR addresses for which the processor will not generate #GP(0).

Implication: For RDMSR, undefined values will be read into EDX:EAX. For WRMSR, undefined

processor behavior may result.

Workaround: Do not use invalid MSR addresses with RDMSR or WRMSR.

54 Specification Update

Errata

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E30. SYSENTER/SYSEXIT Instructions Can Implicitly Load “Null Segment

Selector” to SS and CS Reg isters

Problem: According to the processor specification, attempting to load a null segment selector

into the CS and SS segment registers should generate a General Protection Fault (#GP). Although loading a null segment selector to the other segment registers is allowed, the processor will generate an exception when the segment register holding a null selector is used to access memory.

However, the SYSENTER instruction can implicitly load a null value to the SS segment selector. This can occur if the value in SYSENTER_CS_MSR is between FFF8h and FFFBh when the SYSENTER instruction is executed. This behavior is part of the SYSENTER/SYSEXIT instruction definition; the content of the SYSTEM_CS_MSR is always incremented by 8 before it is loaded into the SS. This operation will set the null bit in the segment selector if a null result is generated, but it does not generate a #GP on the SYSENTER instruction itself. An exception will be generated as expected when the SS register is used to access memory, however.

The SYSEXIT instruction will also exhibit this behavior for both CS and SS when executed with the value in SYSENTER_CS_MSR between FFF0h and FFF3h, or between FFE8h and FFEBh, inclusive.

Implication: These instructions are intended for operating system use. If this erratum occurs (and

the OS does not ensure that the processor never has a null segment selector in the SS or CS segment registers), the processor’s behavior may become unpredictable, possibly resulting in system failure.

Workaround: Do not initialize the SYSTEM_CS_MSR with the values between FFF8h and FFFBh,

FFF0h and FFF3h, or FFE8h and FFEBh before executing SYSENTER or SYSEXIT.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E31. PRELOAD Followed by EXTEST Does Not Load Boundary Scan Data

Problem: According to the IEEE 1149.1 Standard, the EXTEST instruction would use data

“typically loaded onto the latched parallel outputs of boundary-scan shift-register stages using the SAMPLE/PRELOAD instruction prior to the selection of the EXTEST instruction.” As a result of this erratum, this method cannot be used to load the data onto the outputs.

Implication: Using the PRELOAD instruction prior to the EXTEST instruction will not produce

expected data after the completion of EXTEST.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E32. Far Jump to New TSS With D-bit Cleared May Cause System Hang

Specification Update 55

Errata

Problem: A task switch may be performed by executing a far jump through a task gate or to a

new Task State Segment (TSS) directly. Normally, when such a jump to a new TSS occurs, the D-bit (which indicates that the page referenced by a Page Table Entry (PTE) has been modified) for the PTE which maps the location of the previous TSS will already be set, and the processor will operate as expected. However, if the D-bit is clear at the time of the jump to the new TSS, the processor will hang.

Implication: If an OS is used which can clear the D-bit for system pages, and which jumps to a

new TSS on a task switch, then a condition may occur which results in a system hang. Intel has not identified any commercial software which may encounter this condition; this erratum was discovered in a focused testing environment.

Workaround: Ensure that OS code does not clear the D-bit for system pages (including any

pages that contain a task gate or TSS). Use task gates rather than jumping to a new TSS when performing a task switch.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E33. INT 1 Instruction Handler Execution Could Generate a Debug

Exception

Problem: If the processor’s general detect enable flag is set and an explicit call is made to the

interrupt procedure via the INT 1 instruction, the general detect enable flag should be cleared prior to entering the handler. As a result of this erratum, the flag is not cleared prior to entering the handler. If an access is made to the debug registers while inside of the handler, the state of the general detect enable flag will cause a second debug exception to be taken. The second debug exception clears the general detect enable flag and returns control to the handler which is now able to access the debug registers.

Implication: This erratum will generate an unexpected debug exception upon accessing the debug

registers while inside of the INT 1 handler.

Workaround: Ignore the second debug exception that is taken as a result of this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E34. COMISS/UCOMISS May Not Update Eflags Under Certain Conditions

Problem: COMISS/UCOMISS instructions compare the least significant pairs of packed single-

precision floating-point numbers and set the ZF, PF, and CF bits in the EFLAGS register accordingly (the OF, SF, and AF bits are cleared). Under certain conditions when a memory location is loaded into cache, the EFLAGS may not get set.

Implication: The result of the incorrect status of the EFLAGS may range from no effect to an

unexpected application/OS behavior.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

56 Specification Update

Errata

E35. Transmission Error on Cache Read

Problem: During reads of the L2 cache, the processor may use certain L2 cache optimizations

that may result in a data transmission error

Implication: Data corruption caused by this erratum will result in unpredictable system behavior. Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E36. Potential Loss of Data Coherency During MP Data Ownershi p Transfer

Problem: In MP systems, processors may be sharing data in different cache lines, referenced

as line A and line B in the discussion below. When this erratum occurs (with the following example given for a 2-way MP system with processors noted as ‘P0’ and ‘P1’), P0 contains a shared copy of line B in its L1. P1 has a shared copy of Line A. Each processor must manage the necessary invalidation and snoop cycles before that processor can modify and source the results of any internal writes to the other processor.

There exists a narrow timing window when, if P1 requests a coy of line B it may be supplied by P0 in an Exclusive state which allows P1 to modify the contents of the line with no further external invalidation cycles. In this narrow window P0 may also retire instructions that use the original data present before P1 performed the modification

Implication: Multiprocessor or threaded application synchronization, required for low-level data

sharing, that is implemented via operating system provided synchronization constructs are not affected by this erratum. Applications which rely upon the usage of locked semaphores rather than memory ordering are also unaffected. Uniprocessor systems are not affected by this erratum. If the erratum does occur one processor may execute software with the stale data that was present from the previous shared state rather than the data written more recently by another processor.

Workaround: Deterministic barriers beyond which program variables will not be modified can

be achieved via the usage of locked semaphore operations. These should effectively prevent the occurrence of this erratum

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E37. Misaligned Locked Access to APIC Space Results in Hang

Problem: When the processor’s APIC space is accessed with a misaligned locked access a

machine check exception is expected. However, the processor’s machine check architecture is unable to handle the misaligned locked access.

Implication: If this erratum occurs the processor will hang. Typical usage models for the APIC

address space do not use locked accesses. This erratum will not affect systems using such a model.

Workaround: Ensure that all accesses to APIC space are aligned.

Specification Update 57

Errata

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E38. Floating-Point Exception Sig nal May Be Deferred

Problem: A one clock window exists where a pending x87 FP exception that should be signaled

on the execution of a CVTPS2PI, CVTPI2PS, or CVTTPS2PI instruction may be deferred to the next waiting floating-point instruction or instruction that would change MMX™ register state.

Implication: If this erratum occurs the floating-point exception will not be handled as expected. Workaround: Applications that follow Intel programming guidelines (empty all x87 registers

before executing MMX technology instructions) will not be affected by this erratum

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E39. Memory Ordering Based Synchronization May Cause a Livelock

Condition in MP Systems

Problem: In an MP environment, the following sequence of code (or similar code) in two

processors (P0 and P1) may cause them to each enter an infinite loop (livelock condition):

P0 P1

MOV [xyz], EAX (1) wait1: MOV EBX, [abc] (2)

. CMP EBX, val1 (3)

. JNE wait1 (4)

MOV [abc], val1 (6) MOV [abc], val2 (5)

wait0: MOV EBX, [abc] (7)

CMP EBX, val2 (8)

JNE wait0 (9)

NOTE

The EAX and EBX can be any general-purpose register. Addresses [abc] and [xyz] can be any location in memory and must be in the same bank of the L1 cache. Variables “val1” and “val2” can be any integer.

The algorithm above involves processors P0 and P1, each of which use loops to keep them synchronized with each other. P1 is looping until instruction (6) in P0 is globally observed. Likewise, P0 will loop until instruction (5) in P1 is globally observed.

58 Specification Update

Errata

The P6 architecture allows for instructions (1) and (7) in P0 to be dispatched to the L1 cache simultaneously. If the two instructions are accessing the same memory bank in the L1 cache, the load (7) will be given higher priority and will complete, blocking instruction (1).

Instructions (8) and (9) may then execute and retire, placing the instruction pointer back to instruction (7). This is due to the condition at the end of the “wait0” loop being false. The livelock scenario can occur if the timing of the wait0 loop execution is such that instruction (7) in P0 is ready for completion every time that instruction (1) tries to complete. Instruction (7) will again have higher priority, preventing the data ([xyz]) in instruction (1) from being written to the L1 cache. This causes instruction (6) in P0 to not complete and the sequence “wait0” to loop infinitely in P0.

A livelock condition also occurs in P1 because instruction (6) in P0 does not complete (blocked by instruction (1) not completing). The problem with this scenario is that P0 should eventually allow for instruction (1) to write its data to the L1 cache. If this occurs, the data in instruction (6) will be written to memory, allowing the conditions in both loops to be true.

Implication: Both processors will be stuck in an infinite loop, leading to a hang condition. Note

that if P0 receives any interrupt, the loop timing will be disrupted such that the livelock will be broken. The system timer, a keystroke, or mouse movement can provide an interrupt that will break the livelock.

Workaround: Use a LOCK instruction to force P0 to execute instruction (6) before instruction

(7).

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E40. System Bus Address Parity Generator May Report Fa l se AERR#

Problem: The processor’s address parity error detection circuit may fail to meet its frequency

timing specification under certain environmental conditions. At the high end of the temperature specification and/or the low end of the voltage range, the processor may report false address parity errors.

Implication: If the system has AERR# drive enabled (bit [3] of the EBL_CR_POWERON resister set

to ‘1’) spurious address detection and reporting may occur. In some system configurations, BINIT# may be asserted on the system bus. This may cause some systems to generate a machine check exception and in others may cause a reboot.

Workaround: Disable AERR# drive from the processor. AERR# drive may be disabled by

clearing bit [3] in the EBL_CR_POWERON register. In addition, if the chipset allows, AERR# drive should be enabled from the chipset and AERR# observation enabled on the processor. AERR# observation on the processor is enabled by asserting A8# on the active-to-inactive transition of RESET#.

Status: For the processor part numbers affected see the “Pentium® III Processor

Identification and Packaging Information” table in the General Information section.

E41. System Bus ECC Not Functional With 2:1 Ratio

Specification Update 59

Errata

Problem: If a processor is underclocked at a core frequency to system bus frequency ratio of

2:1 and system bus ECC is enabled, the system bus ECC detection and correction will negatively affect internal timing dependencies.

Implication: If system bus ECC is enabled, and the processor is underclocked at a 2:1 ratio, the

system may behave unpredictably due to these timing dependencies.

Workaround: All bus agents that support system bus ECC must disable it when a 2:1 ratio is

used.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E42. Processor May Assert DRDY# on a Write With No Data

Problem: When a MASKMOVQ instruction is misaligned across a chunk boundary in a way that

one chunk has a mask of all 0’s, the processor will initiate two partial write transactions with one having all byte enables deasserted. Under these conditions, the expected behavior of the processor would be to perform both write transactions, but to deassert DRDY# during the transaction which has no byte enables asserted. As a result of this erratum, DRDY# is asserted even though no data is being transferred.

Implication: The implications of this erratum depend on the bus agent’s ability to handle this

erroneous DRDY# assertion. If a bus agent cannot handle a DRDY# assertion in this situation, or attempts to use the invalid data on the bus during this transaction, unpredictable system behavior could result

Workaround: A system which can accept a DRDY# assertion during a write with no data will not

be affected by this erratum. In addition, this erratum will not occur if the MASKMOVQ is aligned.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E43. GP# Fault on WRMSR to ROB_CR_BKUPTMPDR6

Problem: Writing a ‘1’ to unimplemented bit(s) in the ROB_CR_BKUPTMPDR6 MSR (offset 1E0h)

will result in a general protection fault (GP#).

Implication: The normal process used to write an MSR is to read the MSR using RDMSR, modify

the bit(s) of interest, and then to write the MSR using WRMSR. Because of this erratum, this process may result in a GP# fault when used to modify the ROB_CR_BKUPTMPDR6 MSR.

Workaround: When writing to ROB_CR_BKUPTMPDR6 all unimplemented bits must be ‘0.’

Implemented bits may be set as ‘0’ or ‘1’ as desired.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E44. Machine Check Exception May Occur Due to Improper Line Eviction in

the IFU

60 Specification Update

Errata

Problem: The Pentium III processor is designed to signal an unrecoverable Machine Check

Exception (MCE) as a consistency checking mechanism. Under a complex set of circumstances involving multiple speculative branches and memory accesses there exists a one cycle long window in which the processor may signal a MCE in the Instruction Fetch Unit (IFU) because instructions previously decoded have been evicted from the IFU. The one cycle long window is opened when an opportunistic fetch receives a partial hit on a previously executed but not as yet completed store resident in the store buffer. The resulting partial hit erroneously causes the eviction of a line from the IFU at a time when the processor is expecting the line to still be present. If the MCE for this particular IFU event is disabled, execution will continue normally.

Implication: While this erratum may occur on a system with any number of Pentium III

processors, the probability of occurrence increases with the number of processors. If this erratum does occur, a machine check exception will result. Note systems that implement an operating system that does not enable the Machine Check Architecture will be completely unaffected by this erratum (e.g., Windows* 95 and Windows 98).

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E45. Performance Counters Include Streaming SIMD Extensions L1

Prefetch

Problem: The processor allows the measurement of the frequency and duration of numerous

different internal and bus related events (see Intel Architecture Software Developer's Manual, Volume 3, for more details). The Streaming SIMD Extension (SSE) architecture provides a mechanism to pre-load data into the L1 cache, bypassing the L2 cache. The number of these L1 pre-loads measured by the performance monitoring logic will incorrectly be included in the count of “L2_LINES_IN” (24H) events and “L2_LINES_OUT” (26H) events.

Implication: If application software is run which utilizes the SSE L1 prefetch feature, the count of

“L2_LINES_IN” (24H) and “L2_LINES_OUT” (26H) will read a value that is greater than the correct value.

Workaround: The correct value of “L2_LINES_IN” and “L2_LINES_OUT” may be calculated by

subtracting the value of the “MMX_PRE_MISS” (4BH) from each of these registers.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E46. Snoop Request May Cause DBSY# Hang

Problem: A small window of time exists in which a snoop request originating from a bus agent

to a processor with one or more outstanding memory transactions may cause the processor to assert DBSY# without issuing a corresponding bus transaction, causing the processor to hang (livelock). The exact circumstances are complex, and include the relative timing of internal processor functions with the snoop request from a bus agent

Specification Update 61

Errata

Implication: This erratum may occur on a system with any number of processors. However, the

probability of occurrence increases with the number of processors. If this erratum does occur, the system will hang with DBSY# asserted. At this point, the system requires a hard reset.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E47. Lower Bits of SMRAM SMBASE Register Cannot Be Written With an

ITP

Problem: The System Management Base (SMBASE) register (7EF8H) stores the starting

address of the System Management RAM (SMRAM). This register is used by the processor when it is in System Management Mode (SMM), and its contents serve as the memory base for code execution and data storage. The 32-bit SMBASE register can normally be programmed to any value. When programmed with an In-Target Probe (ITP), however, any attempt to set the lower 11 bits of SMBASE to anything other than zeros via the WRMSR instruction will cause the attempted write to fail.

Implication: When set via ITP, any attempt to relocate SMRAM space must be made with 2-KB

alignment.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E48. Task Switch Caused by Page Fault May Cause Wrong PTE and PDE

Access Bit to Be Set

Problem: If an operating system executes a task switch via a Task State Segment (TSS), and

the TSS is wholly or partially located within a clean page (A and D bits clear) and the GDT entry for the new TSS is either misaligned across a cache line boundary or is in a clean page, the accessed and dirty bits for an incorrect page table/directory entry may be set.

Implication: An operating system which uses hardware task switching (or hardware task

management) may encounter this erratum. The effect of the erratum depends on the alignment of the TSS and ranges from no anomalous behavior to unexpected errors.

Workaround: The operating system could align all TSSs to be within page boundaries and set

the A and D bits for those pages to avoid this erratum. The operating system may alternately use software task management.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E49. Unsynchronized Cross-Modifying Code Operations Can Cause

Unexpected Instruction Execution Results

62 Specification Update

Errata

Problem: The act of one processor, or system bus master, writing data into a currently

executing code segment of a second processor with the intent of having the second processor execute that data as code is called cross-modifying code (XMC). XMC that does not force the second processor to execute a synchronizing instruction, prior to execution of the new code, is called unsynchronized XMC.

Software using unsynchronized XMC to modify the instruction byte stream of a processor can see unexpected instruction execution from the processor that is executing the modified code.

Implication: In this case, the phrase "unexpected execution behavior" encompasses the

generation of most of the exceptions listed in the Intel Architecture Software Developer's Manual Volume 3: System Programming Guide, including a General Protection Fault (GPF). In the event of a GPF the application executing the unsynchronized XMC operation would be terminated by the operating system.

Workaround: In order to avoid this erratum, programmers should use the XMC synchronization

algorithm as detailed in the Intel Architecture Software Developer's Manual Volume 3: System Programming Guide, Section 7.1.3.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E50. Processor Will Erroneously Report a BIST Failure

Problem: If the processor performs BIST at power-up, the EAX register is normally cleared

(0H) when the processor passes. The processor will erroneously report a non-zero value (signaling a BIST failure) even if BIST passes.

Implication: The processor will incorrectly signal an error after BIST is performed. Workaround: The system BIOS should ignore the BIST results in the EAX register. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E51. Noise Sensitivity Issue on Processor SMI# Pin

Problem: Post silicon characterization has demonstrated a greater than expected sensitivity to

noise on the processor's SMI# input, which may result in spurious SMI# interrupts.

Implication: BIOS/SMM code that is capable of handling spurious SMI events will report a spurious

SMI#, but should not be negatively impacted by this erratum. Systems whose BIOS code cannot handle spurious SMI events may fail, resulting in a system hang or other anomalous behavior.

Spurious SMI# interrupts should be controlled on the system board regardless of BIOS implementation.

Workaround: Possible workarounds that may reduce or eliminate the occurrence of the

spurious SMI# interrupts include:

1. Use a lower effective pull-up resistance on the SMI# pin. This resistor must meet the specifications of the component driving the SMI# signal.

Specification Update 63

Errata

2. Externally condition the SMI# signal prior to providing it to the processor's SMI# pin.

These workarounds should be evaluated on a design-by-design basis.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E52. Limitation On Cache Line ECC Detection and Correction

Problem: ECC can detect and correct up to four single-bit ECC errors per cache line. However,

the processor will only detect and correct one single-bit ECC error per cache line. While all ECC errors will be detected, multiple single bit errors will be incorrectly reported as uncorrectable double bit errors, rather than correctable single bit errors.

Implication: The processor may report fewer single bit ECC errors and more double bit ECC errors

than previous processors.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E53. L2_LD and L2_M_LINES_OUTM Performance-Monitoring Counters Do

Not Work

Problem: The L2_LD (29H) and L2_M_LINES_OUTM (27H) Performance-Monitoring counters

are used to monitor L2 cache line activity. These counters incorrectly count their respective events.

Implication: These counters will report incorrect data. Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E54. IFU/DCU Deadlock May Cause System Hang

Problem: An internal deadlock situation may occur in systems with multiple bus agents, with a

failure signature such that a processor either asserts DBSY# without issuing the corresponding data, or fails to respond to a snoop request from another bus agent. Should this erratum occur, the affected processor ceases code execution and the system will hang.

The specific circumstances surrounding the occurrence of this erratum are:

1. A locked operation to the Data Cache Unit (DCU) is in process.

2. A snoop occurs, but cannot complete due to the ongoing locked operation.

3. The presence of the snoop prevents pending Instruction Fetch Unit (IFU) requests from completing.

64 Specification Update

Errata

4. The IFU requests are periodically restarted.

The continued IFU restart attempts create additional DCU snoops, which prevent the in-process locked operation from completing, keeping the DCU locked.

Implication: The system may hang Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E55. L2_DBUS_BUSY Performance Monitoring Counter Will Not Count

Writes

Problem: The L2_DBUS_BUSY (22H) performance monitoring counter is intended to count the

number of cycles during which the L2 data bus is in use. For some steppings of the processor, the L2_DBUS_BUSY counter will not be incremented during write cycles and therefore will only reflect the number of L2 data bus cycles resulting from cache reads

Implication: The L2_DBUS_BUSY event counts only L2 read cycles. Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E56. Incorrect Sign May Occur On X87 Result Due To Indefinite QNaN

Result From Streaming SIMD Extensions Multiply

Problem: It is possible that a negative sign bit may be incorrectly applied to the result of an

X87 floating-point operation if it is closely preceded by a Streaming SIMD Extensions (SSE) multiply operation. In order for this erratum to occur, the Streaming SIMD Extensions multiply operation must result in an Indefinite Quiet Not-a-Number (QNaN). Operations such as multiplying zero by infinity will result in an Indefinite QNaN result.

Implication: If this erratum occurs, the result of an X87 floating-point instruction, which should be

positive, will instead be negative.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E57. Deadlock May Occur Due To Illegal-Instruction/Page-Miss

Combination

Problem: Intel's 32-bit Instruction Set Architecture (ISA) utilizes most of the available op-code

space; however some byte combinations remain undefined and are considered illegal instructions. Intel processors detect the attempted execution of illegal instructions and signal an exception. This exception is handled by operating system and/or application software.

Specification Update 65

Errata

Under a complex set of internal and external conditions involving illegal instructions, a deadlock may occur within the processor. The necessary conditions for the deadlock involve:

1. Execution of the illegal instruction.

2. Two-page table walks occur within a narrow timing window coincident with the illegal instruction.

Implication: The illegal instructions involved in this erratum are unusual and invalid byte

combinations that are not useful to application software or operating systems. These combinations are not normally generated in the course of software programming, nor are such sequences known by Intel to be generated in commercially available software and tools. Development tools (compilers, assemblers) do not generate this type of code sequence, and will normally flag such a sequence as an error. If this erratum occurs, the processor deadlock condition will occur and result in a system hang. Code execution cannot continue without a system RESET.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E58. MASKMOVQ Instruction Interaction with String Operation May Cause

Deadlock

Problem: Under the following scenario, combined with a specific alignment of internal events,

the processor may enter a deadlock condition:

1. A store operation completes, leaving a write-combining (WC) buffer partially

filled.

2. The target of a subsequent MASKMOVQ instruction is split across a cache line.

3. The data in (2) above results in a hit to the data in the WC buffer in (1).

Implication: If this erratum occurs, the processor deadlock condition will occur and result in a

system hang. Code execution cannot continue without a system RESET.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E59. MOVD, CVTSI2SS, or PINSRW Following Zeroing Instruction Can

Cause Incorrect Result

Problem: An incorrect result may be calculated after the following circumstances occur:

1. A register has been zeroed with either a SUB reg, reg instruction, or an XOR reg, reg instruction.

2. A value is moved with sign extension into the same register’s lower 16 bits; or a signed integer multiply is performed to the same register’s lower 16 bits.

66 Specification Update

Errata

3. The register is then copied to an MMX™ technology register using the MOVD, or converted to single precision floating-point and moved to an MMX technology register using the CVTSI2SS instruction prior to any other operations on the signextended value, or inserted into an MMX™ technology register using the PINSRW instruction.

Specifically, the sign may be incorrectly extended into bits 16-31 of the MMX technology register. In the case of the PINSRW instruction, a non-zero value could be loaded into the MMX™ technology register. This erratum only affects the MMX™ technology register.

This erratum only occurs when the following three steps occur in the order shown. This erratum may occur with up to 63 (39 for Pre-CPUID 0x6BX) intervening instructions that do not modify the sign-extended value between steps 2 and 3.

1. XOR EAX, EAX

or SUB EAX, EAX

2. MOVSX AX, BL

or MOVSX AX, byte ptr <memory address> or MOVSX AX, BX

or MOVSX AX, word ptr <memory address> or IMUL BL (AX implicit, opcode F6

/5)

or IMUL byte ptr <memory address> (AX implicit, opcode F6 /5) or IMUL AX, BX

(opcode 0F AF /r)

or IMUL AX, word ptr <memory address> (opcode 0F AF /r) or IMUL AX, BX, 16

(opcode 6B /r ib)

or IMUL AX, word ptr <memory address>, 16 (opcode 6B /r ib) or IMUL AX, 8

(opcode 6B /r ib)

or IMUL AX, BX, 1024 (opcode 69 /r iw)

or IMUL AX, word ptr <memory address>, 1024 (opcode 69 /r iw)

or IMUL AX, 1024 (opcode 69 /r iw) or CBW

3. MOVD MM0, EAX or CVTSI2SS MM0, EAX

Note that the values for immediate byte/words are merely representative (i.e., 8, 16,

1024) and that any value in the range for the size is affected. Also, note that this

erratum may occur with “EAX” replaced with any 32-bit general-purpose register, and “AX” with the corresponding 16-bit version of that replacement. “BL” or “BX” can be

Specification Update 67

Errata

replaced with any 8-bit or 16-bit general-purpose register. The CBW and IMUL (opcode F6 /5) instructions are specific to the EAX register only.

In the above example, EAX is forced to contain 0 by the XOR or SUB instructions. Since the four types of the MOVSX or IMUL instructions and the CBW instruction only modify bits 15:8 of EAX by sign extending the lower 8 bits of EAX, bits 31:16 of EAX should always contain 0. This implies that when MOVD or CVTSI2SS copies EAX to MM0, bits 31:16 of MM0 should also be 0. In certain scenarios, bits 31:16 of MM0 are not 0, but are replicas of bit 15 (the 16th bit) of AX. This is noticeable when the value in AX after the MOVSX, IMUL or CBW instruction is negative (i.e., bit 15 of AX is a 1).

When AX is positive (bit 15 of AX is 0), MOVD or CVTSI2SS will produce the correct answer. If AX is negative (bit 15 of AX is 1), MOVD or CVTSI2SS may produce the right answer or the wrong answer, depending on the point in time when the MOVD or CVTSI2SS instruction is executed in relation to the MOVSX, IMUL or CBW instruction.

The PINSRW instruction can fail to correctly load a zero when used with a partial register zeroing instruction (SUB or XOR):

1. mov di, 0FFFF8914h

2. xor eax, eax

3. add ax, di

4. xor ah, ah

5. pinsrw mm1, eax, 00h

In this case, the programmer expects mm1 to contain 0014h in it’s least significant word. This erratum would cause MM1 to contain 8914h. The number of intervening instructions between steps 4 and 5 is the same as noted in the sign extension example above between steps 2 and 3.

Implication: The effect of incorrect execution will vary from unnoticeable, due to the code

sequence discarding the incorrect bits, to an application failure.

Workaround: There are two possible workarounds for this erratum:

1. Rather than using the MOVSX-MOVD/CVTSI2SS, IMUL-MOVD/CVTSI2SS or CBW-

MOVD/CVTSI2SS pairing to handle one variable at a time, use the sign extension capabilities (PSRAW, etc.) within MMX technology for operating on multiple variables. This will also result in higher performance.

2. Insert another operation that modifies or copies the sign-extended value between

the MOVSX/IMUL/CBW instruction and the MOVD or CVTSI2SS instruction as in the example below:

XOR EAX, EAX (or SUB EAX, EAX)

68 Specification Update

Errata

MOVSX AX, BL (or other MOVSX, other IMUL or CBW instruction)

*MOV EAX, EAX

MOVD MM0, EAX or CVTSI2SS MM0, EAX

3. Avoid using a sub or xor to zero a partial register prior to the use of any of these

three instructions. Instead, use a mov immediate (e.g. “mov ah, 0h”).

*Note: MOV EAX, EAX is used here in a generic sense. Again, EAX can be substituted with any 32-bit register.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E60. FLUSH# Assertio n Following STPCL K# May Prevent CPU Clo cks From

Stopping

Problem: If FLUSH# is asserted after STPCLK# is asserted, the cache flush operation will not

occur until after STPCLK# is de-asserted. Furthermore, the pending flush will prevent the processor from entering the Sleep state, since the flush operation must complete prior to the processor entering the Sleep state.

Implication: Following SLP# assertion, processor power dissipation may be higher than expected.

Furthermore, if the source to the processor’s input bus clock (BCLK) is removed, normally resulting in a transition to the Deep Sleep state, the processor may shutdown improperly. The ensuing attempt to wake up the processor will result in unpredictable behavior and may cause the system to hang.

Workaround: For systems that use the FLUSH# input signal and Deep Sleep state of the

processor, ensure that FLUSH# is not asserted while STPCLK# is asserted.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E61. Intermittent Failure to Assert ADS# Duri ng Processor Pow er- On

Problem: Under a system specific set of initial parametric conditions, a very small number of

Pentium® III processors (CPUID 068xh) may be susceptible to entering an internal test mode during processor power-on. The symptom of this test mode is a failure to assert ADS# during a processor power-on.

Implication: On susceptible platforms, when power is applied to the processor, there is a

possibility that the processor will occasionally enter the test mode rather than initiate a system boot sequence.

A subsequent processor Power-Off then Power-On cycle should remove the processor from this test mode, allowing normal processor operation to resume. The following workaround also may reduce the occurrence of the failure condition:

Specification Update 69

Errata

SC242-based platform designs in which VTT leads the processor input voltage may reduce the occurrence of the erratum by connecting SC242 pin B20 (RESERVED) to pin B9 (VTT).

PGA370-based platform designs in which VTT leads the processor input voltage can reduce the occurrence of the erratum by connecting pin G37 (RESERVED) to motherboard VTT or short the PGA370 socket pin G37 to AH20 or G35 (both VTT).

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E62. Floating-Point Exception C ond i t i on May b e Deferred

Problem: A floating-point instruction that causes a pending floating-point exception (ES=1) is

normally signaled by the processor on the next waiting FP/MMX™ technology instruction. In the following set of circumstances, the exception may be delayed or the FSW register may contain a wrong value:

1. The excepting floating-point instruction is followed by an instruction that accesses memory across a page (4-Kbyte) boundary or its access results in the update of a page table dirty/access bit.

2. The memory accessing instruction is immediately followed by a waiting floatingpoint or MMX technology instruction.

3. The waiting floating-point or MMX technology instruction retires during a onecycle window that coincides with a sequence of internal events related to instruction cache line eviction.

Implication: The floating-point exception will not be signaled until the next waiting floating-

point/MMX technology instruction. Alternatively it may be signaled with the wrong TOS and condition code values. This erratum has not been observed in any commercial software applications.

Workaround: None identified Status: For the stepping affected see the Summary of Changes at the beginning of this

section.

E63. THERMTRIP# May Not be Asserted as Specified

Problem: THERMTRIP# is a signal on the Pentium® III processor that is asserted when the

core reaches a critical temperature during operation as detailed in the processor specification. The Pentium® III processor may not assert THERMTRIP# until a much higher temperature than the one specified is reached.

Implication: The THERMTRIP# feature is not functional on the Pentium® III processor. Note that

this erratum can only occur when the processor is running with a TPLATE temperature over the maximum specification of 75° C

Workaround: Avoid operation of the Pentium III processor outside of the thermal specifications

defined by the processor specifications.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

70 Specification Update

Errata

E64. Cache Line Reads May Result in Eviction of Invalid Data

Problem: A small window of time exists in which internal timing conditions in the processor

cache logic may result in the eviction of an L2 cache line marked in the invalid state.

Implication: There are three possible implications of this erratum:

1. The processor may provide incorrect L2 cache line data by evicting an invalid line.

2. A BNR# (Block Next Request) stall may occur on the system bus.

3. Should a snoop request occur to the same cache line in a small window of time the processor may incorrectly assert HITM#. It is then possible for an infinite snoop stall to occur should another processor respond (correctly) to the snoop request with HIT#. In order for this infinite snoop stall to occur, at least three agents must be present, and the probability of occurrence increases with the number of processors. Should 2 or 3 occur, the processor will eventually assert BINIT# (if enabled) with an MCA error code indicating a ROB time-out. At this point, the system requires a hard reset.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E65. Snoop probe during FLUSH# could cause L2 to be left in shared state

Problem: During a L2 FLUSH operation using the FLUSH# pin, it is possible that a read request

from a bus agent or other processor to a valid line will leave the line in the Shared state (S) instead of the Invalid state (I) as expected after flush operation. Before the FLUSH operation is completed, another snoop request to invalidate the line from another agent or processor could be ignored, again leaving the line in the Shared state.

Implication: Current desktop and mid range server systems have no mechanism to assert the

flush pin and hence are not affected by this erratum. A high-end server system that does not suppress snoop traffic before the assertion of the FLUSH# pin may cause a line to be left in an incorrect cache state.

Workaround: Affected systems (those capable of asserting the FLUSH# pin) should prevent

snoop activity on the front side bus until invalidation is completed after asserting FLUSH#, or use a WBINVD instruction instead of asserting the FLUSH# pin in order to flush the cache.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E66. Livelock May Occur Due to IFU Line Eviction

Problem: Following the conditions outlined for erratum E44, if the instruction that is currently

being executed from the evicted line must be restarted by the IFU, and the IFU receives another partial hit on a previously executed (but not as yet completed) store that is resident in the store buffer, then a livelock may occur.

Specification Update 71

Errata

Implication: If this erratum occurs, the processor will hang in a live lock-situation, and the system

will require a reset to continue normal operation.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E67. Selector for the LTR/LLDT Register May Get Corrupted

Problem: The internal selector portion of the respective register (TR, LDTR) may get corrupted,

if during a small window of LTR or LLDT system instruction execution, the following sequence of events occur:

1. Speculative write to a segment register that might follow the LTR or LLDT instruction

2. The read segment descriptor of LTR/LLDT operation spans a page (4 Kbytes) boundary; or causes a page fault

Implication: Incorrect selector for LTR, LLDT instruction could be used after a task switch. Workaround: Software can insert a serializing instruction between the LTR or LLDT instruction

and the segment register write.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E68. INIT Does Not Clear Global Entries in the TLB

Problem: INIT may not flush a TLB entry when:

1. The processor is in protected mode with paging enabled and the page global enable flag is set (PGE bit of CR4 register)

2. G bit for the page table entry is set

3. TLB entry is present in TLB when INIT occurs

Implication: Software may encounter unexpected page fault or incorrect address translation due

to a TLB entry erroneously left in TLB after INIT.

Workaround: Write to CR3, CR4 or CR0 registers before writing to memory early in BIOS code

to clear all the global entries from TLB.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E69. VM Bit will be Cleared on a Double Fault Handler

Problem: Following a task switch to a Double Fault Handler that was initiated while the

processor was in virtual-8086 (VM86) mode, the VM bit will be incorrectly cleared in EFLAGS.

72 Specification Update

Errata

Implication: When the OS recovers from the double fault handler, the processor will no longer be

in VM86 mode

Workaround: None identified Status: For the steppings affected see the Summary of Errata at the beginning of this section

E70. Memory Aliasing with Inconsistent A and D Bits may Cause Processor

Deadlock

Problem: In the event that software implements memory aliasing by having two Page Directory

Entries (PDEs) point to a common Page Table Entry (PTE) and the Accessed and Dirty bits for the two PDEs are allowed to become inconsistent the processor may become deadlocked

Implication: This erratum has not been observed with commercially available software Workaround: Software that needs to implement memory aliasing in this way should manage

the consistency of the Accessed and Dirty bits.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E71. Use of Memory Aliasing with Inconsistent Memory Type May Cause

System Hang

Problem: Software that implements memory aliasing by having more than one linear addresses

mapped to the same physical page with different cache types may cause the system to hang. This would occur if one of the addresses is non-cacheable used in code segment and the other a cacheable address. If the cacheable address finds its way in instruction cache, and non-cacheable address is fetched in IFU, the processor may invalidate the non-cacheable address from the fetch unit. Any micro-architectural event that causes instruction restart will expect this instruction to still be in fetch unit and lack of it will cause system hang

Implication: This erratum has not been observed with commercially available software Workaround: Although it is possible to have a single physical page mapped by two different

linear addresses with different memory types, Intel has strongly discouraged this practice as it may lead to undefined results. Software that needs to implement memory aliasing should manage the memory type consistency

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E72. Processor may Report Invalid TSS Fault Instead of Double Fault

During Mode C Paging

Problem: When an operating system executes a task switch via a Task State Segment (TSS)

the CR3 register is always updated from the new task TSS. In the mode C paging, once the CR3 is changed the processor will attempt to load the PDPTRs. If the CR3 from the target task TSS or task switch handler TSS is not valid then the new PDPTR

Specification Update 73

Errata

will not be loaded. This will lead to the reporting of invalid TSS fault instead of the expected Double fault

Implication: Operating systems that access an invalid TSS may get invalid TSS fault instead of a

Double fault.

Workaround: Software needs to ensure any accessed TSS is valid. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E73. Machine Check Exception May Occur When Interleaving Code

Between Different Memory Typ es

Problem: A small window of opportunity exists where code fetches interleaved between

different memory types may cause a machine check exception. A complex set of micro-architectural boundary conditions is required to expose this window.

Implication: Interleaved instruction fetches between different memory types may result in a

machine check exception. The system may hang if machine check exceptions are disabled. Intel has not observed the occurrence of this erratum while running commercially available applications or operating systems.

Workaround: Software can avoid this erratum by placing a serializing instruction between code

fetches between different memory types.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E74. Wrong ESP Register Values During a Fault in VM86 Mode

Problem: At the beginning of the IRET instruction execution in VM86 mode, the lower 16 bits of

the ESP register are saved as the old stack value. When a fault occurs, these 16 bits are moved into the 32-bit ESP, effectively clearing the upper 16 bits of the ESP.

Implication: This erratum has not been observed to cause any problems with commercially

available software.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E75. APIC ICR Write May Cause Interrupt Not to be Sent When ICR

Delivery Bit Pending

Problem: If the APIC ICR (Interrupt Control Register) is written with a new interrupt command

while the Delivery Status bit from a previous interrupt command is set to '1' (Send Pending), the interrupt message may not be sent out by the processor.

Implication: This erratum will cause an interrupt message not to be sent, potentially resulting in

system hang.

74 Specification Update

Errata

Workaround: Software should always poll the Delivery Status bit in the APIC ICR and ensure

that it is '0' (Idle) before writing a new value to the ICR.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E76. High Temperature and Low Supply Voltage Operation May Result in

Incorrect Processor Operation

Problem: When operating at the high temperature, low supply voltage corner of the processor

specification, if there is a store pending in the processor’s fill buffer, and simultaneously a load operation misses the L1 cache but results in a hit to the L2 cache, then it is possible that incorrect data may be returned to satisfy the load operation.

Implication: When this erratum is encountered, unpredictable software behavior may occur. It can

be seen from the table of affected steppings that this erratum is constrained to a single stepping and is only possible in processors operating at frequencies of 933MHz and above and is not present in all of those processors. Application of the workaround will prevent occurrence of the erratum in all processors of that stepping.

Workaround: It is possible for BIOS code to contain a workaround for this erratum. Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E77. During Boundary Scan, BCLK not Sampled High When SLP# is

Asserted Low

Problem: During boundary scan, BCLK is not sampled high when SLP# is asserted low. Implication: Boundary scan results may be incorrect when SLP# is asserted low. Workaround: Do not use boundary scan when SLP# is asserted low. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E78. Incorrect Assertion of THERMTRIP# Signal

Problem: The internal control register bit responsible for operation of the Thermtrip circuit

functionality may power up in a non-initialized state. As a result, THERMTRIP# may be incorrectly asserted during de-assertion of RESET# at nominal operating temperatures. When THERMTRIP# is asserted as a result of this erratum, the processor may shut down internally and stop execution but in few cases continue to execute.

Implication: This issue can lead to intermittent system power-on boot failures. The occurrence

and repeatability of failures is system dependent, however all systems and processors are susceptible to failure. In addition, the processor may fail to stop execution during the event of a valid THERMTRIP# assertion resulting in the potential for permanent processor damage

Specification Update 75

Errata

Workaround: To prevent the risk of power-on boot failures or catastrophic thermal failures, a

platform workaround is required. The system must provide a rising edge on the TCK signal during the power-on sequence that meets all of the following requirements:

• Rising edge occurs after Vcc_core is valid and stable

• Rising edge occurs before or at the de-assertion of RESET#

• Rising edge occurs after all Vref input signals are at valid voltage levels

• TCK input meets the Vih min and max spec as mentioned in EMTS

Specific workaround implementations may be platform specific. The following examples have been tested as acceptable workaround implementations.

In addition, the example workaround circuits shown do not support production baseboard test methodologies that require the use of the processor JTAG/TAP port. Alternative workaround solutions must be found if such test capability is required.

Pentium® III Processor with 512KB L2 Cache Platforms Workaround

2.5V

330 ohm

PWRGD

Assumes the inputs to the CPU_PWRGD are open collector signals that are Wire-ANDed together

150 ohm

CPU1

0 ohm

680 ohm

TCK

CPU2

PWRGD

76 Specification Update

Errata

• The example workaround circuit assumes that the PWRGD inputs into the processors are open collector. Tying the PWRGD inputs together in a Wired-AND fashion allows each processor to receive PWRGD at the same time but at the latter of the 2 separate PWRGD assertions. If separation of the PWRGD inputs to each processor is required, extra circuitry will be required.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E79. Processor Might not Exit Sleep State Properly Upon De-assertion of

CPUSLP# Signal

Problem: If the processor enters a sleep state upon assertion of CPUSLP# signal, and if the

core to system bus multiplier is an odd bus fraction, then the processor may not resume from the CPU sleep state upon the de-assertion of CPUSLP# signal.

Implication: This erratum may result in a system hang during a resume from CPU sleep state. Workaround: It is possible to workaround this in BIOS by not asserting CPUSLP# for power

management purposes

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E80. The Instruction Fetch Unit (IFU) May Fetch Instructions Based Upon

Stale CR3 Data after a Write to CR3 Register

Problem: Under a complex set of conditions, there exists a one-clock window following a write

to the CR3 register wherein it is possible for the iTLB fill buffer to obtain a stale page translation based on the stale CR3 data. This stale translation will persist until the next write to the CR3 register, the next page fault or execution of a certain class of instructions including CPUID or IRETD with privilege level change.

Implication: The wrong page translation could be used leading to erroneous software behavior. Workaround: Operating systems that are potentially affected can add a second write to the

CR3 register.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E81. Unde r So me Complex Conditio ns, th e Ins truc tions in the Shadow of a

JMP FAR may be Unintentionally Executed and Retired

Problem: If all of the following events happen in sequence it is possible for the system or

application to hang or to execute with incorrect data.

1. The execution of an instruction, with an OPCODE that requires the processor to

stall the issue of micro-instructions in the flow from the microcode sequence logic block to the instruction decode block. (a StallMS condition)

2. Less than 63 (39 for Pre-CPUID 0x6BX) micro-instructions later, the execution of a

mispredictable branch instruction. (Jcc, LOOPcc, RET Near, CALL Near Indirect, JMP ECX=0, or JMP Near Indirect)

Specification Update 77

Errata

3. The conditional branch in event (2) is mispredicted, and furthermore the

mispredicted path of execution must result in either an ITLB miss, or an Instruction Cache miss. This needs to briefly stall the issue of micro-instructions again immediately after the conditional branch until that branch prediction is corrected by the jump execution block. (a 2nd StallMS condition)

4. Along the correct path of execution, the next instruction must contain a 3rd

StallMS condition at a precisely aligned point in the execution of the instruction. ( CLTS, POPSS, LSS, or MOV to SS)

5.A JMP FAR instruction must execute within the next 63 micro-instructions (39 Pre-

CPUID 0x6BX) The intervening micro-instructions must not have any events or faults.

When the instruction from event (2) retires, the StallMS condition within the event (5) instruction fails to operate correctly, and instructions in the shadow of the JMP FAR instruction could be unintentionally executed.

Implication: Occurrence of this erratum could lead to erroneous software behavior. Intel has not

identified any commercially available software which may encounter this condition; this erratum was discovered in a focused test environment. One of the four instructions that are required to trigger this erratum, CLTS, is a privileged instruction that is only executed by an operating system or driver code. The remaining three instructions, POPSS, LSS, and MOV to SS, are executed infrequently in modern 32-bit application code.

Workaround: None identified at this time Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E82. Processor Does not Flag #GP on Non-zero Write to Certain MSRs

Problem: When a non-zero write occurs to the upper 32 bits of SYSENTER_EIP_MSR or

SYSENTER_ESP_MSR, the processor should indicate a general protection fault by flagging #GP. Due to this erratum, the processor does not flag #GP.

Implication: The processor unexpectedly does not flag #GP on a non-zero write to the upper 32

bits of SYSENTER_EIP_MSR or SYSENTER_ESP_MSR. No known commercially available operating system has been identified to be affected by this erratum.

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E83. IFU/BSU Deadlock May Cause System Hang

Problem: A lockable instruction with memory operand that spans across two pages may, given

some rare internal conditions, hang the system.

Implication: When this erratum occurs, the system may hang. Intel has not observed this erratum

with any commercially available software or system.

Workaround: Lockable data should always be contained in a single page.

78 Specification Update

Errata

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E84. REP MOVS Operation in Fast String Mode Continues in That Mode

When Crossing into a Page with a Different Memory Type

Problem: A fast “REP MOVS” operation will continue to be handled in fast mode when the string

operation crosses a page boundary into an Uncacheable (UC) memory type. Also if the fast string operation crosses a page boundary into a WC memory region, the processor will not self snoop the WC memory region. This may eventually result in incorrect data for the WC portion of the operation if those cache lines were previously cached as WB (through aliasing) and modified.

Implication: String elements should be handled by the processor at the native operand size in UC

memory. In the event that the WB to WC aliasing case occurs and incorrect data is written to memory, the end result would vary from benign to operating system or application failure. Intel has not observed either aspects of this erratum in commercially available software.

Workaround: Software operating within Intel’s recommendation will not require WB and WC

memory aliased to the same physical address.

Status: For the steppings affected, see the Summary Table of Changes.

E85. The FXSAVE, STOS or MOVS Instructions May Cause a Store Ordering

Violation When Data Crosses a Page with a UC Memory Type

Problem: If the data from an FXSAVE, STOS or MOVS instruction crosses a page boundary from

WB to UC memory type and this instruction is immediately followed by a second instruction that also issues a store to memory, the final data stores from both instructions may occur in the wrong order

Implication: The impact of this store ordering behavior may vary from normal software execution

to potential software failure. Intel has not observed this erratum in commercially available software.

Workaround: FXSAVE, STOS or MOVS data must not cross page boundary from WB to UC

memory type.

Status: For the steppings affected, see the Summary Table of Changes.

E86. POPF and POPFD Instructions that Set the Trap Flag Bit May Cause

Unpredictabl e Processor Behav i or

Problem: In some rare cases, POPF and POPFD instructions that set the Trap Flag (TF) bit in

the EFLAGS register (causing the processor to enter Single-Step mode) may cause unpredictable processor behavior.

Implication: Single step operation is typically enabled during software debug activities, not during

normal system operation

Specification Update 79

Errata

Workaround: There is no workaround for single step operation in commercially available

software. For debug activities on custom software, the POPF and POPFD instructions could be immediately followed by a NOP instruction to facilitate correct execution.

Status: For the steppings affected, see the Summary Table of Changes.

E87. Code Segment Limit Violation May Occur on 4 Gigabyte Limit Check

Problem: Code Segment limit violation may occur on 4 Gigabyte limit check when the code

stream wraps around in a way that one instruction ends at the last byte of the segment and the next instruction begins at 0x0.

Implication: This is a rare condition that may result in a system hang. Intel has not observed this

erratum with any commercially available software, or system.

Workaround: Avoid code that wraps around segment limit. Status: For the steppings affected, see the Summary Table of Changes.

E88. FST Instruction with Numeric and Null Segment Exceptions May

Cause Gen eral Protectio n Faults to be Missed and FP Linear Address (FLA) Mismatch

Problem: FST instruction combined with numeric and null segment exceptions may cause

General Protection Faults to be missed and FP Linear Address (FLA) mismatch.

Implication: This is a rare condition that may result in a system hang. Intel has not observed this

erratum with any commercially available software, or system.

Workaround: None identified. Status: For the steppings affected, see the Summary Table of Changes.

E89. Code Segment (CS) is incorrect on SMM Handler when SMBASE is not

Aligned

Problem: With SMBASE being relocated to a non-aligned address, during SMM entry the CS can

be improperly updated which can lead to an incorrect SMM handler.

Implication: This is a rare condition that may result in a system hang. Intel has not observed this

erratum with any commercially available software, or system.

Workaround: Align SMBASE to 32K byte. Status: For the steppings affected, see the Summary Table of Changes.

E90. Page with PAT (Page Attribute Table) Set to USWC (Uncacheable

Speculative Write Combine) While Associated MTRR (Memory Type Range Register) is UC (Uncacheable) May Consolidate to UC

Problem: A page whose PAT memory type is USWC while the relevant MTRR memory type is

UC, the consolidated memory type may be treated as UC (rather than WC as specified in IA-32 Intel® Architecture Software Developer's Manual).

80 Specification Update

Errata

Implication: When this erratum occurs, the memory page may be as UC (rather than WC). This

may have a negative performance impact.

Workaround: None identified. Status: For the steppings affected, see the Summary Table of Changes.

E91. Under Certain Conditions LTR (Load Task Register) Instruction May

Result in System Hang

Problem: An LTR instruction may result in a system hang if all the following conditions are met:

1. Invalid data selector of the TR (Task Register) resulting with either #GP (General Protection Fault) or #NP (Segment Not Present Fault).

2. GDT (Global Descriptor Table) is not 8-Bytes aligned. GDT (Global Descriptor Table) is not 8-Bytes aligned.

3. Data BP (breakpoint) is set on cache line containing the descriptor data.

Implication: This erratum may result in system hang if all conditions have been met. This

erratum has not been observed in commercial operating systems or software. For performance reasons, GDT is typically aligned to 8-Bytes.

Workaround: Software should align GDT to 8-Bytes. Status: For the steppings affected, see the Summary Table of Changes.

E92. Loading from Memory Type USWC (Uncacheable Speculative Write

Combine) May Get Its Data Internally Forwarded From a Previous Pending Store

Problem: A load from memory type USWC may get its data internally forwarded from a

pending store. As a result, the expected load may never be issued to the external bus.

Implication: When this erratum occurs, a USWC Load request may be satisfied without being

observed on the external bus. There are no known usage models where this behavior results in any negative side-effects.

Workaround: Do not use memory type USWC for memory that has read side effects. Status: For the steppings affected, see the Summary Table of Changes.

E93. FXSAVE after FNINIT Without an Intervening FP (Floating Point)

Instruction May Save Uninitialized Values for FDP (x87 FPU Instruction Operand (Data) Pointer Offset) and FDS (x87 FPU Instruction Operand (Data) Pointer Selector)

Problem: An FXSAVE after FNINIT without an intervening FP instruction may save uninitialized

values for FDP and FDS.

Implication: When this erratum occurs, the values for FDP/FDS in the FXSAVE structure may

appear to be random values. These values will be initialized by the first FP

Specification Update 81

Errata

instruction executed after the FXRSTOR that restore the saved floating point state. Any FP instruction with memory operand will initialize FDP/FDS. Intel has not observed this erratum with any commercially available software.

Workaround: After an FINIT, do not expect the FXSAVE memory image to be correct, until at

least one FP instruction with a memory operand has been executed.

Status: For the steppings affected, see the Summary Table of Changes.

E94. FST P ( Flo a tin g P o in t S to r e) I ns t ru c tio n Un de r Ce r ta in Co n ditio n s M a y

Result In Erroneously Setting a Valid Bit on an FP (Floating Point) Stack Register

Problem: An FSTP instruction with an PDE/PTE (Page Directory Entry/Page Table Entry) A/D bit

update followed by user mode access fault due to a code fetch to a page that has supervisor only access permission may result in erroneously setting a valid bit of an FP stack register. The FP top of stack pointer is unchanged.

Implication: This erratum may cause an unexpected stack overflow Workaround: User mode code should not count on being able to recover from illegal accesses

to memory regions protected with supervisor only access when using FP instructions.

Status: For the steppings affected, see the Summary Table of Changes.

E95. Invalid Entries in Page-Directory-Pointer-Table Register (PDPTR)

May Cause General Protection (#GP) Exception if the Reserved Bits are Set to One

Problem: Invalid entries in the Page-Directory-Pointer-Table Register (PDPTR) that have the

reserved bits set to one may cause a General Protection (#GP) exception.

Implication: Intel has not observed this erratum with any commercially available software. Workaround: Do not set the reserved bits to one when PDPTR entries are invalid. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E96. Writing the Local Vector Table (LVT) when an Interrupt is Pending

May Cause an Unexpected Interrup t

Problem: If a local interrupt is pending when the LVT entry is written, an interrupt may be

taken on the new interrupt vector even if the mask bit is set.

Implication: An interrupt may immediately be generated with the new vector when a LVT entry is

written, even if the new LVT entry has the mask bit set. If there is no Interrupt Service Routine (ISR) set up for that vector the system will GP fault. If the ISR does not do an End of Interrupt (EOI) the bit for the vector will be left set in the in-service register and mask all interrupts at the same or lower priority.

Workaround: Any vector programmed into an LVT entry must have an ISR associated with it,

even if that vector was programmed as masked. This ISR routine must do an EOI to

82 Specification Update

Errata

clear any unexpected interrupts that may occur. The ISR associated with the spurious vector does not generate an EOI, therefore the spurious vector should not be used when writing the LVT

Status: For the steppings affected, see the Summary Tables of Changes.

E97. The Processor May Report a #TS Instead of a #GP Fault

Problem: A jump to a busy TSS (Task-State Segment) may cause a #TS (invalid TSS

exception) instead of a #GP fault (general protection exception).

Implication: Operation systems that access a busy TSS may get invalid TSS fault instead of a #GP

fault. Intel has not observed this erratum with any commercially available software.

Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Changes.

E98. A Write to an APIC Register Sometimes May Appear to Have Not

Occurred

Problem: With respect to the retirement of instructions, stores to the uncacheable memory-

based APIC register space are handled in a non-synchronized way. For example if an instruction that masks the interrupt flag, e.g. CLI, is executed soon after an uncacheable write to the Task Priority Register (TPR) that lowers the APIC priority, the interrupt masking operation may take effect before the actual priority has been lowered. This may cause interrupts whose priority is lower than the initial TPR, but higher than the final TPR, to not be serviced until the interrupt enabled flag is finally set, i.e. by STI instruction. Interrupts will remain pending and are not lost.

Implication: In this example the processor may allow interrupts to be accepted but may delay

their service.

Workaround: This non-synchronization can be avoided by issuing an APIC register read after

the APIC register write. This will force the store to the APIC register before any subsequent instructions are executed. No commercial operating system is known to be impacted by this erratum..

Status: For the steppings affected, see the Summary Tables of Changes.

E99. Using 2M/4M Pages When A20M# Is Asserted May Result in

Incorrect Address Translations

Problem: An external A20M# pin if enabled forces address bit 20 to be masked (forced to zero)

to emulates real-address mode address wraparound at 1 megabyte. However, if all of the following conditions are met, address bit 20 may not be masked.

• paging is enabled

• a linear address has bit 20 set

• the address references a large page

• A20M# is enabled

Specification Update 83

Errata

Implication: When A20M# is enabled and an address references a large page the resulting

translated physical address may be incorrect. This erratum has not been observed with any commercially available operating system.

Workaround: Operating systems should not allow A20M# to be enabled if the masking of

address bit 20 could be applied to an address that references a large page. A20M# is normally only used with the first megabyte of memory

Status: For the steppings affected, see the Summary Tables of Change.

E100. Values for LBR/BTS/BTM will be Incorrect after an Exit from SMM

Problem: After a return from SMM (System Management Mode), the CPU will incorrectly update

the LBR (Last Branch Record) and the BTS (Branch Trace Store), hence rendering their data invalid. The corresponding data if sent out as a BTM on the system bus will also be incorrect. Note: This issue would only occur when one of the 3 above mentioned debug support facilities are used.

Implication: The value of the LBR, BTS, and BTM immediately after an RSM operation should not

be used.

Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Change.

E101. INIT Does Not Clear Global Entries in the TLB

Problem: INIT may not flush a TLB entry when:

1. The processor is in protected mode with paging enabled and the page global enable flag is set (PGE bit of CR4 register)

2. G bit for the page table entry is set

3. TLB entry is present in TLB when INIT occurs.

Implication: Software may encounter unexpected page fault or incorrect address translation due

to a TLB entry erroneously left in TLB after INIT.

Workaround: Write to CR3, CR4 (setting bits PSE, PGE or PAE) or CR0 (setting bits PG or PE)

registers before writing to memory early in BIOS code to clear all the global entries from TLB.

Status: For the steppings affected, see the Summary Table of Changes.

E102. REP MOVS/STOS Executing with Fast Strings Enabled and Crossing

Page Boundaries with Inconsistent Memory Types may use an Incorrect Data Size or Lead to Memory-Ordering Violations

Problem: Under certain conditions as described in the Software Developers Manual section

“Out-of-Order Stores For String Operations in Pentium 4, Intel Xeon, and P6 Family Processors” the processor performs REP MOVS or REP STOS as fast strings. Due to

84 Specification Update

Errata

this erratum fast string REP MOVS/REP STOS instructions that cross page boundaries from WB/WC memory types to UC/WP/WT memory types, may start using an incorrect data size or may observe memory ordering violations.

Implication: Upon crossing the page boundary the following may occur, dependent on the new

page memory type:

• UC the data size of each write will now always be 8 bytes, as opposed to the original data size.

• WP the data size of each write will now always be 8 bytes, as opposed to the original data size and there may be a memory ordering violation.

• WT there may be a memory ordering violation.

Workaround: Software should avoid crossing page boundaries from WB or WC memory type to

UC, WP or WT memory type within a single REP MOVS or REP STOS instruction that will execute with fast strings enabled

Status: For the steppings affected, see the Summary Tables of Changes

E103. The BS Flag in DR6 May be Set for Non-Single-Step #DB Exception

Problem: DR6 BS (Single Step, bit 14) flag may be incorrectly set when the TF (Trap Flag, bit

8) of the EFLAGS Register is set, and a #DB (Debug Exception) occurs due to one of

the following:

• DR7 GD (General Detect, bit 13) being bit set;

• INT1 instruction;

• Code breakpoint

the DR6 BS (Single Step, bit 14) flag may be incorrectly set.

Implication: The BS flag may be incorrectly set for non-single-step #DB exception. Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Changes

E104. Fault on ENTER Instruction May Result in Unexpected Values on

Stack Frame

Problem: The ENTER instruction is used to create a procedure stack frame. Due to this

erratum, if execution of the ENTER instruction results in a fault, the dynamic storage area of the resultant stack frame may contain unexpected values (i.e. residual stack data as a result of processing the fault).

Implication: Data in the created stack frame may be altered following a fault on the ENTER

instruction. Please refer to "Procedure Calls For Block-Structured Languages" in IA32 Intel® Architecture Software Developer’s Manual, Vol. 1, Basic Architecture, for information on the usage of the ENTER instructions. This erratum is not expected to occur in ring 3. Faults are usually processed in ring 0 and stack switch occurs when

Specification Update 85

Errata

transferring to ring 0. Intel has not observed this erratum on any commercially available software.

Workaround: None identified. Status: For the steppings affected, see the Summary Tables of Changes

E105. Unaligned Accesses to Paging Structures May Cause the Processor to

Hang

Problem: when an unaligned access is performed on paging structure entries, accessing a

portion of two different entries simultaneously, the processor may live lock.

Implication: When this erratum occurs, the processor may live lock causing a system hang. Workaround: Do not perform unaligned accesses on paging structure entries. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E106. INVLPG Operation for Large (2M/4M) Pages May be Incomplete

under Certain Condit i ons

Problem: The INVLPG instruction may not completely invalidate Translation Look-aside Buffer

(TLB) entries for large pages (2M/4M) when both of the following conditions exist:

• Address range of the page being invalidated spans several Memory Type Range Registers (MTRRs) with different memory types specified

• INVLPG operation is preceded by a Page Assist Event (Page Fault (#PF) or an access that results in either A or D bits being set in a Page Table Entry (PTE))

Implication: Stale translations may remain valid in TLB after a PTE update resulting in

unpredictable system behavior. Intel has not observed this erratum with any commercially available software.

Workaround: Software should ensure that the memory type specified in the MTRRs is the same

for the entire address range of the large page.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E107. P age Access Bit May be Set Prio r to Signaling a Code Segment Limit

Fault

Problem: If code segment limit is set close to the end of a code page, then due to this erratum

the memory page Access bit (A bit) may be set for the subsequent page prior to general protection fault on code segment limit.

Implication: When this erratum occurs, a non-accessed page which is present in memory and

follows a page that contains the code segment limit may be tagged as accessed

86 Specification Update

Errata

Workaround: Erratum can be avoided by placing a guard page (non-present or non-executable

page) as the last page of the segment or after the page that includes the code segment limit.

Status: For the steppings affected see the Summary of Changes at the beginning of this

section

E108. EFLAGS, CR0, CR4 and the EXF4 Signal May be Incorrect after

Shutdown

Problem: When the processor is going into shutdown due to an RSM inconsistency failure,

EFLAGS, CR0 and CR4 may be incorrect. In addition the EXF4 signal may still be asserted. This may be observed if the processor is taken out of shutdown by NMI#.

Implication: A processor that has been taken out of shutdown may have an incorrect EFLAGS,

CR0 and CR4. In addition the EXF4 signal may still be asserted.

Workaround: None identified Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

E109. Performance Monitoring Event FP_MMX_TRANS_TO_MMX May Not

Count Some Transitions

Problem: Performance Monitor Event FP_MMX_TRANS_TO_MMX (Event CCH, Umask 01H)

counts transitions from x87 Floating Point (FP) to MMX™ instructions. Due to this erratum, if only a small number of MMX instructions (including EMMS) are executed immediately after the last FP instruction, an FP to MMX transition may not be counted.

Implication: The count value for Performance Monitoring Event FP_MMX_TRANS_TO_MMX may be

lower than expected. The degree of undercounting is dependent on the occurrences of the erratum condition while the counter is active. Intel has not observed this erratum with any commercially available software.

Workaround: None identified. Status: For the steppings affected see the Summary of Changes at the beginning of this

section.

Specification Update 87

Specification Changes

The Specification Changes listed in this section apply to the following documents:

• Intel® Pentium® III Processor for the SC242 at 450 MHz to 1.13 GHz datasheet

• Intel® Pentium® III Processor for the PGA370 Socket up to 1.13 GHz datasheet

• Intel® Pentium® III Processor on 0.13 Micron Process Up to 1.40 GHz Electrical,

Mechanical and Thermal Specification

All Specification Changes will be incorporated into a future version of the appropriate Pentium III processor documentation.

Specification Changes

There are no Specification Changes for this Month

88 Specification Update

Specification Clarifications

The Specification Clarifications listed in this section apply to the following documents:

• Pentium® III Processor for the SC242 at 450MHz to 1.13GHz datasheet

• Pentium® III Processor for the PGA370 Socket at 500MHz to 1.13 GHz datasheet

• Pentium® III Processor on 0.13 micron process Up to 1.40 GHz datasheet

• Pentium® III Processor with 512KB L2 Cache datasheet

• Intel® 64 and IA-32 Architectures Software Developer's Manual Volume 1, 2a,

2b, 3a and 3b: System Programming Guide

All Specification Clarifications will be incorporated into a future version of the appropriate Pentium III processor documentation.

E1. Specification Clarification with respect to Time Stamp Counter

In the “Debugging and Performance Monitoring” chapter (Sections 15.8, 15.10.9 and

15.10.9.3) of the IA-32 Intel® Architecture Software Developer’s Manual Volume 3:

System Programming Guide, the Time Stamp Counter definition has been updated to include support for the future processors. This change will be incorporated in the next revision of the IA-32 Intel® Architecture Software Developer’s Manual.

15.8 Time-Stamp Counter

The IA-32 architecture (beginning with the Pentium processor) defines a time-stamp counter mechanism that can be used to monitor and identify the relative time occurrence of processor events. The counter’s architecture includes the following components:

• TSC flag — A feature bit that indicates the availability of the time-stamp counter.

The counter is available in an IA-32 processor implementation if the function CPUID.1:EDX.TSC[bit 4] = 1.

• IA32_TIME_STA MP_COUNTER MSR (called TSC MSR in P6 family and Pentium

processors) — The MSR used as the counter.

• RDTSC instruction — An instruction used to read the time-stamp counter.

• TSD flag — A control register flag is used to enable or disable the time-stamp

counter (enabled if CR4.TSD[bit 2] = 1).

The time-stamp counter (as implemented in the P6 family, Pentium, Pentium M, Pentium 4, and Intel Xeon processors) is a 64-bit counter that is set to 0 following a RESET of the processor. Following a RESET, the counter will increment even when the

Specification Update 89

Specification Clarifications

processor is halted by the HLT instruction or the external STPCLK# pin. Note that the assertion of the external DPSLP# pin may cause the time-stamp counter to stop.

Members of the processor families increment the time-stamp counter differently:

• For Pentium M processors (family [06H], models [09H, 0DH]); for Pentium 4

processors, Intel Xeon processors (family [0FH], models [00H, 01H, or 02H]); and for P6 family processors: the time-stamp counter increments with every internal processor clock cycle. The internal processor clock cycle is determined by the current core-clock to bus-clock ratio. Intel® SpeedStep® technology transitions may also impact the processor clock.

• For Pentium 4 processors, Intel Xeon processors (family [0FH], models [03H and

higher]): the time-stamp counter increments at a constant rate. That rate may be set by the maximum core-clock to bus-clock ratio of the processor or may be set by the frequency at which the processor is booted. The specific processor configuration determines the behavior. Constant TSC behavior ensures that the duration of each clock tick is uniform and supports the use of the TSC as a wall clock timer even if the processor core changes frequency. This is the architectural behavior moving forward.

NOTE

To determine average processor clock frequency, Intel recommends the use of Performance Monitoring logic to count processor core clocks over the period of time for which the average is required. See Section 15.10.9 and Appendix A in this manual for more information.

The RDTSC instruction reads the time-stamp counter and is guaranteed to return a monotonically increasing unique value whenever executed, except for a 64-bit counter wraparound. Intel guarantees that the time-stamp counter will not wraparound within 10 years after being reset. The period for counter wrap is longer for Pentium 4, Intel Xeon, P6 family, and Pentium processors.

Normally, the RDTSC instruction can be executed by programs and procedures running at any privilege level and in virtual-8086 mode. The TSD flag allows use of this instruction to be restricted to programs and procedures running at privilege level

0. A secure operating system would set the TSD flag during system initialization to

disable user access to the time-stamp counter. An operating system that disables user access to the time-stamp counter should emulate the instruction through a user-accessible programming interface.

The RDTSC instruction is not serializing or ordered with other instructions. It does not necessarily wait until all previous instructions have been executed before reading the counter. Similarly, subsequent instructions may begin execution before the RDTSC instruction operation is performed.

The RDMSR and WRMSR instructions read and write the time-stamp counter, treating the time-stamp counter as an ordinary MSR (address 10H). In the Pentium 4, Intel Xeon, and P6 family processors, all 64-bits of the time-stamp counter are read using RDMSR (just as with RDTSC). When WRMSR is used to write the time-stamp counter on processors before family [0FH], models [03H, 04H]: only the low order 32-bits of the time-stamp counter can be written (the high-order 32 bits are cleared to 0). For family [0FH], models [03H, 04H]: all 64 bits are writeable.

90 Specification Update

Specification Clarifications

Counting Clocks

The count of cycles, also known as clockticks, forms a the basis for measuring how long a program takes to execute. Clockticks are also used as part of efficiency ratios like cycles per instruction (CPI). Processor clocks may stop ticking under circumstances like the following:

• The processor is halted when there is nothing for the CPU to do. For example, the

processor may halt to save power while the computer is servicing an I/O request. When Hyper-Threading Technology is enabled, both logical processors must be halted for performance-monitoring counters to be powered down.

• The processor is asleep as a result of being halted or because of a power-

management scheme. There are different levels of sleep. In the some deep sleep levels, the time-stamp counter stops counting.

There are three ways to count processor clock cycles to monitor performance. These are:

• Non-halted clockticks — Measures clock cycles in which the specified logical

processor is not halted and is not in any power-saving state. When HyperThreading Technology is enabled, this these ticks can be measured on a perlogical-processor basis.

• Non-sleep clockticks — Measures clock cycles in which the specified physical

processor is not in a sleep mode or in a power-saving state. These ticks cannot be measured on a logical-processor basis.

• Time-stamp counter — Some processor models permit clock cycles to be

measured when the physical processor is not in deep sleep (by using the timestamp counter and the RDTSC instruction). Note that such ticks cannot be measured on a per-logical-processor basis. See Section 10.8 for detail on processor capabilities.

The first two methods use performance counters and can be set up to cause an interrupt upon overflow (for sampling). They may also be useful where it is easier for a tool to read a performance counter than to use a time stamp counter (the timestamp counter is accessed using the RDTSC instruction).

For applications with a significant amount of I/O, there are two ratios of interest:

• Non-halted CPI — Non-halted clockticks/instructions retired measures the CPI

for phases where the CPU was being used. This ratio can be measured on a logical-processor basis when Hyper-Threading Technology is enabled.

• Nominal CPI — Time-stamp counter ticks/instructions retired measures the CPI

over the duration of a program, including those periods when the machine halts while waiting for I/O.

15.10.9.3

Incrementing the Time-Stamp Counter

The time-stamp counter increments when the clock signal on the system bus is active and when the sleep pin is not asserted. The counter value can be read with the RDTSC instruction.

Specification Update 91

Specification Clarifications

The time-stamp counter and the non-sleep clockticks count may not agree in all cases and for all processors. See Section 10.8 for more information on counter operation.

92 Specification Update

Documentation Changes

Docume ntation Cha ng e s

The Documentation Changes listed in this section apply to the following documents:

• Pentium® III Processor for the SC242 at 450MHz to 1.13GHz datasheet

• Pentium® III Processor for the PGA370 Socket at 500MHz to 1.13 GHz datasheet

• Pentium® III Processor on 0.13 micron process datasheet

• Pentium® III Processor with 512KB L2 Cache datasheet

All Documentation Changes will be incorporated into a future version of the appropriate Pentium III processor documentation.

Note: Documentation changes for Intel® 64 and IA-32 Architectures Software Developer’s Manual volumes 1, 2a, 2b, 3a and 3b are posted in a separate document "Intel® 64 and IA-32 Architectures Software Developer’s Manual Documentation Changes". This Document has been posted to http://developer.intel.com/.

There are no Documentation Changes for this Month

Specification Update 93

Intel SL3QA - Pentium III 550 MHz Processor, SL6BY - Pentium III Processor 1.4GHz Tualatin 512KB/133/370, Pentium III Specification

Specifications and Main Features

Frequently Asked Questions

User Manual