Intel® Xeon™ Processor with
512 KB L2 Cache and Intel®
E7500 Chipset Platform
Design Guide
March 2002
Document Number: 298649-002
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 descriptions 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 the m.
The Intel® E7500 chipset and processors in the Intel® Xeon processor family may contain design defects or errors known as errata which may cause the product
to deviate from published specifi catio ns. C urr en t chara cter ized err ata are avail a ble on req ues t.
Contact your local Intel sales office or your distributor to obtain the latest specifications and before placing your product order.
2
I
C is a two-wire communications bus/protocol developed by Philips. SMBus is a subset of the I2C bus/protocol and was developed by Intel.
Implementations of the I2C bus/protocol may require licenses from various entities, including Philips Electronics N.V. and North American Philips
Corporation.
Alert on LAN and Alert on LAN2 are a result of the Intel-IBM Advanced Manageability Alliance and a trademark of IBM
Copies of documents which have an ordering number and are referenced in this document, or other Intel literature, may be obtained from:
Intel Corporation
www.intel.com
or call 1-800-548-4725
Intel, Pentium, Intel Xeon, Intel Netburst and the Intel logo are trademarks or registered trademarks of Intel Corporation or its subsidiaries in the United
Added: Section 12.5.4; New P64H2 Power Sequencing
Requirement
Updated Schematics to reflect changes identified above.
March 2002
14Design Guide
Introduction
Introduction1
The Intel® Xeon™ Processor with 512 KB L2 Cache and Intel® E7500 Chipset Platform Design
Guide documents In tel’s design recommendations fo r systems based on the Intel
Processor with 512 KB L2 Cache and the E7500 chipset. In addition to providing motherboard
design recommendations such as layout and routing guidelines, this document addresses system
design issues such as power delivery.
Carefully follow the design information, board schematics, debug recommendations, and system
checklists provided in this document. These design guidelines have been developed to ensure
maximum flexibility for board designers while reducing the risk of board related issues.
Note that the guidelines recommended in this document are based on experience and simulation
work done at Intel while developing Intel Xeon processor with 512 KB L2 cache / E7500 chipsetbased systems. This work is ongoing, and the recommendations are subject to change.
Board designers may use the associated I ntel schematics as a referen ce. While the schematics cover
a specific design implementation, the core schematics remain the same for most E7500 chipsetbased platforms. The schematic set provides a reference schematic for each E7500 chipset
component as well as common motherboard options. Additional flexibility is possible through
other permutations of these options and components.
1.1Reference Documentation
Note:For the latest revision and documentation number, contact your appropriate field representative.
This section defines conventions and terminology used throughout the design guide.
Convention/TerminologyDescription
Aggressor A network that transmits a coupled signal to another network.
AGTL+ The Xeon™ processor family system bus uses a bus technology called AGTL+,
Asynchronous GTL+Xeon processors do not utilize CMOS voltage levels on any signals that connect
Bus AgentA component or group of components that, when combined, represent a single
Core PowerCore power refers to a power rail that is on only during full-power operation.
CrosstalkThe reception on a victim network of a signal imposed by aggressor network(s)
Derived powerA derived power rail is any power rail that is generated from another power rail
Dual Processor (DP)Used to specify a system configuration using two processors.
Electromagnetic
Compatibility (EMC)
Electromagnetic
Interference (EMI)
or Assisted Gunning Transceiver Logic. AGTL+ buffers are open-drain, and
require pull-up resistors to provide the high logic level and termination. AGTL+
output buffers differ from GTL+ buffers with the addition of an active pMOS pullup transistor to assist the pull-up resistors during the first clock of a low-to-high
voltage transition.
to the processor. As a result, legacy input signals such as A20M#, IGNNE#,
INIT#, LINT0/INTR, LINT1/NMI, PWRGOOD, SMI#, SLP#, and STPCLK# utilize
GTL+ input buffers. Legacy output signals (FERR# and IERR#) and non-AGTL+
signals (THERMTRIP# and PROCHOT#) also utilize GTL+ output buffers. All of
these signals follow the same DC requirements as AGTL+ signals, however the
outputs are not actively driven high (during a logical 0 to 1 transition) by the
processor (the major difference between GTL+ and AGTL+). These signals do
not have setup or hold time specifications in relation to BCLK[1:0], and are
therefore referred to as “Asynchronous GTL+ Signals”. However, all of the
Asynchronous GTL+ signals are required to be asserted for at least two BCLKs
in order for the processor to recognize them.
load on the AGTL+ bus.
These power rails are on when the active-low PSON signal is asserted to the
power supply. The core power rails that are distributed directly from the power
supply are: +12 V, +5 V, and +3.3 V.
through inductive and capacitive coupling between the networks.
• Backward Crosstalk – Coupling that creates a signal in a victim network that
travels in the opposite direction as the aggressor’s signal.
• Forward Crosstalk – Coupling that creates a signal in a victim network that
travels in the same direction as the aggressor’s signal.
• Even Mode Crosstalk – Coupling from a signal or multiple aggressors when
all the aggressors switch in the same direction that the victim is switching.
• Odd Mode Crosstalk – Coupling from a signal or multiple aggressors when
all the aggressors switch in the opposite direction that the victim is switching.
using an on-board voltage regulator. For example, +2.5 V is derived from a +5 V
power rail using a voltage regulator.
The successful operation of electronic equipment in its intended electromagnetic
environment.
Electromagnetic radiation from an electrical source that interrupts the normal
function of an electronic device.
Introduction
Design Guide17
Introduction
Convention/TerminologyDescription
Flight TimeFlight time is a term in the timing equation that includes the signal propagation
Full-powerDuring full-power operation, all components on the motherboard remain
GTLREFReference voltage for AGTL+ input pins.
Inter-Symbol
Interference (ISI)
NetworkThe network is the trace of a Printed Circuit Board (PCB) that completes an
OvershootThe maximum voltage observed for a signal at the device pad, measured with
PadThe electrical contact point of a semiconductor die to the package substrate. A
PinThe contact point of a component package to the traces on a substrate, such as
Power-Good“Power-Good,” “PWRGOOD,” or “CPUPWRGOOD” (an active high signal)
Power RailsA power supply has five power rails: +12 V, –12 V, +5 V, +3.3 V, and +5 VSB. In
RingbackThe voltage to which a signal changes after reaching its maximum absolute
System BusThe System Bus is the bus which connects the processor to the platform.
Setup WindowThe time between the beginning of Setup to Clock (TSU_MIN) and the arrival of
delay, any effects the system has on the Tco of the driver, plus any adjustments
to the signal at the receiver needed to guarantee the setup time of the receiver.
More precisely, flight time is defined as:
• The time difference between a signal at the input pin of a receiving agent
crossing the switching voltage (adjusted to meet the receiver manufacturer’s
conditions required for AC timing specifications; i.e., ringback, etc.) and the
output pin of the driving agent crossing the switching voltage when the driver
is driving a test load used to specify the driver’s AC timings.
• Maximum and Minimum Flight Time – Flight time variations are caused by
many different parameters. The more obvious causes include variation of
the board dielectric constant, changes in load condition, crosstalk, power
noise, variation in termination resistance, and differences in I/O buffer
performance as a function of temperature, voltage, and manufacturing
process. Some less obvious causes include effects of Simultaneous
Switching Output (SSO) and packaging effects.
• Maximum flight time is the largest acceptable flight time that a network
experiences under all conditions.
• Minimum flight time is the smallest acceptable flight time that a network
experiences under all conditions.
powered. Note that full-power operation includes both the full-on operating state,
and the S1 (processor stop-grant) state.
The effect of a previous signal (or transition) on the interconnect delay. For
example, when a signal is transmitted down a line, and the reflections due to the
transition have not completely dissipated, the following data transition launched
onto the bus is affected. ISI is dependent upon frequency, time delay of the line,
and the reflection coefficient at the driver and receiver. ISI can impact both timing
and signal integrity.
electrical connection between two or more components.
respect to VCC.
pad is only observable in simulations.
the motherboard. Signal quality and timings can be measured at the pin.
indicates that all of the system power supplies and clocks are stable.
PWRGOOD should go active a predetermined time after system voltages are
stable and should go inactive as soon as any of these voltages fail their
specifications.
addition to these power rails from the power supply, several other power rails are
created by voltage regulators on the Reference Board.
value. Ringback may be caused by reflections, driver oscillations, or other
transmission line phenomena.
a valid clock edge. This window may be different for each type of bus agent in
the system.
18Design Guide
Convention/TerminologyDescription
Introduction
Simultaneous Switching
Output (SSO)
Standby Power RailStandby power is supplied by the power supply during times when the system is
StubThe branch from the bus trunk terminating at the pad of an agent.
TrunkThe main connection, excluding interconnect branches, from one end agent pad
UndershootThe minimum voltage extending below VSS observed for a signal at the device
VCC_CPUVCC_CPU is the core power for the processor. The system bus is terminated to
VictimA network that receives a coupled crosstalk signal from another network is called
VRM 9.1 “VRM 9.1” refers to the Voltage Regulator Module specification for the Xeon
Effects which are differences in electrical timing parameters and degradation in
signal quality caused by multiple signal outputs simultaneously switching voltage
levels in the opposite direction from a single signal or in the same direction.
These are called odd mode and even mode switching, respectively. This
simultaneous switching of multiple outputs creates higher current swings that
may cause additional propagation delay (“push-out”) or a decrease in
propagation delay (“pull-in”). These SSO effects may impact the setup and/or
hold times and are not always taken into account by simulations. System timing
budgets should include margin for SSO effects.
powered down. The purpose is to maintain functions that always need to be
enabled, such as the date and time-of-day within the BIOS. The power supply
provides a +5 VSB power rail.
to the other end agent pad.
pad.
VCC_CPU.
the victim network.
processor. It is a DC-DC converter module that supplies the required voltage and
current to a single processor.
1.3System Overview
The E7500 chipset is Intel’s first generation server chipset designed for use with the Xeon
processor. The architecture of the chipset provides the performance and feature-set required for
dual-processor based severs in the entry- level and mid-range, front -end and g eneral-purp ose server
market segments. A new chipset component interconnect, the Hub Interface 2.0 (HI2.0), is
designed into the E7500 chipset to provide more efficient communication between chipset
components for high-speed I/O . Each HI2.0 prov ides 1.066 GB/s I/O bandwidt h. The E7500 MC H
has three HI2.0 connectio ns , d eliv eri ng 3.2 G B /s b andw id th fo r h i gh- sp eed I /O, which can be used
for PCI/PCI-X bridges. The system bus, used to connect the processor with the E7500 chipset,
utilizes a 400 MHz transfer rate for data transfers, delivering 3.2 GB/s. The E7500 chipset
architecture supports a 144-bit wide, 200 MHz DDR memory interface also capable of transferring
data at 3.2 GB/s.
In addition to these performance features, E7500 chipset-based platforms also provide the RASUM
(Reliability, Availability, Serviceability, Usability, and Manageability) features required for entrylevel and mid-range servers. These features include: Chipkill* technology ECC for memory, ECC
for all high-performance I/O, out-of-band manageability through SMBus target interfaces on all
major components, memory scrubbing and auto-initialization, processor thermal monito ring, and
hot-plug PCI. For a complete list of the features on this platform, refer to the component datasheets
listed inSection 1.1.
Design Guide19
Introduction
1.3.1Intel® Xeon™ Processor with 512 KB L2 Cache
The Intel Xeon processor with 512 KB L2 cache is the second generation of microprocessors
targeted for severs and workstations using the Intel
processor delivers performance levels that are sign ificantly hig her th an previou s g eneratio ns of IA32 processors. The E7500 chipset supports all speeds of the Intel Xeon processor with 512 KB L2
cache.
Table 1-2. Intel
L2 Cache512 KB
Data Bus Transfer Rate3.2 GB/s
Multi-Processor Support1–2 CPUs
Manageability Features
Package603-pin micro-PGA
Operating Voltage1.50 V
Unless otherwise noted, the term “processor” refers to the Xeon processor.
The Xeon processor includes the following advanced microarchitecture features:
• Hyper Pipelined Technology.
• Advanced Dynamic Execution.
• Execution Trace Cache.
®
NetBurst™ microarchitecture. The Xeon
®
Xeon™ Processor with 512 KB L2 Cache Feature Set Overview
The Intel Xeon processor system bus utilizes a split-transaction, deferred reply protocol similar to
that of the Intel
with the Pentium III Xeon processor bus. The system bus uses source-synchronous transfer of
address and data to improve performance and enables addressing at 2X the system bus frequency
of 100 MHz and data transfers at 4X the system bus frequency of 100 MHz. This allows the
processors to transfer data at 3.2 GB/s.
The Xeon processor provides manageability features consistent with Intel
processors. These features include the Processor Information ROM, the OEM EEPROM, and the
processor thermal sensor; all of which are accessed through the System Management Bus
(SMBus). The Processor Information ROM is a 128-byte read-only device that incorporates Intel
processor specific data. The OEM EEPROM, also known as the “scratchpad EEPROM,” is a
128-byte read/write EEPROM in which an OEM may program system specific data. The thermal
sensor monitors the temperature of the processor die.
®
Pentium III Xeon™ processor bus, however the system bus is not compatible
®
Pentium® III Xeon™
20Design Guide
1.3.2Intel® E7500 Chipset
Introduction
The E7500 chipset consists of three major components: the Intel® E7500 Memory Controller Hub
(referred to throughout this document as the MCH), the Intel
(hereafter referred to as ICH3-S), and th e Intel
P64H2). The chipset components communicate via hub interfaces (HIs). The MCH provides four
hub interface connections: one for the ICH3-S and three for high-speed I/O using 82870P2 P64H2
components. The hub interfaces are point-to-po int and therefore on ly support two agents (the MCH
plus one I/O device). Therefore, the system supports a total of three P64H2s.
®
82870P2 PCI/ PCI-X 64-bit Hub 2 (abbreviated to
1.3.2.1Intel® E7500 Memory Controller Hub (MCH)
The MCH is a 1005-ball FC-BGA package and contains the following functionality:
• System Bus Features:
— Supports dual processors at 100 MHz (x4 transfers).
— System bus bandwidth of 3.2 GB/s (400 MHz).
— Supports 36-bit system bus addressing model.
— 12 deep in-order queue, 2 deep defer queue.
• Memory Bus Features:
— 144-bit wide, DDR-200 memory interface with memory bandwidth of 3.2 GB/s.
— Supports x72, ECC, registered DDR-200 DIMMs using 64-Mb, 128-Mb, 256-Mb and
512-Mb DRAMs.
— Supports a maximum of 16 GB of memory.
— Supports Single 4-bit Error Correct, Double 4-bit Error Detect (S4EC/D4ED) Chipkill
technology ECC (x4 Chipkill technology).
— Supports up to 32 simultaneous open pages.
®
82801CA I/O Controller Hub 3-S
• I/O Features:
— Provides HI1.5 connection for ICH3-S (Hub Interface A):
- 266 MB/s point-to-point connection for ICH3-S with parity protection.
- 8-bit wide, 66 MHz base clock, 4X data transfer.
- Parallel termination mode for longer trace lengths.
— Supports C0, C1, C2, S0, S1, S4, and S5 power states. (Does not support C3, C4, S2,
and S3).
Design Guide21
Introduction
1.3.2.2I/O Controller Hub 3 (Intel® ICH3-S)
The I/O Controller Hub (ICH3-S) provides the legacy I/O subsystem for E7500 chipset-based
platforms. Additionally, it integrates many advanced I/O functions. The ICH3-S includes the
following features:
• Provides HI1.5 Connection to MC H:
— 266 MB/s point-to-point connection for ICH3-S with parity protection.
— 8-bit wide, 66 MHz base clock, 4X data transfer.
— Parallel termination mode for longer trace lengths.
— 64-bit inbound addressing, 32-bit outbound addressing.
• 2 channel Ultra ATA/100 bus master IDE controller.
The P64H2 provides PCI/PCI-X, high-performance I/O capability on E7500 chipset based
platforms. Each P64H2 component includes:
• 16-bit, HI2.0 Connection to MCH:
— 1 GB/s point-to-point connection for I/O bridges with ECC protection.
— 16-bit wide, 66 MHz base clock, 8X data transfer.
— Parallel termination mode for longer trace lengths.
— 64-bit inbound addressing, 32-bit outbound addressing.
• Two Independent, 64-bit PCI/PCI-X Interfaces:
— PCI-X Specification, Revision 1.0a compliant.
— PCI Local Bus Specification, Revision 2.2 compliant.
— PCI-PCI Bridge Architecture Specification, Revision 1.1 compliant.
— PCI Hot Plug Specification, Revision 1.1 compliant.
— One PCI Hot Plug Controller (PHPC) per PCI/PCI-X interface.
— One IOxAPIC per PCI/PCI-X Interface (16 external, 8 internal interrupts).
— SMBus target for access to all internal PCI registers.
22Design Guide
1.3.3Bandwid th Summary
Table 1-3 describes the clock maximum speed, sample rate, and bandwidth for each of the
interfaces in the E7500 chipset based platform.
Table 1-3. Platform Maximum Bandwidth Summary
Introduction
Interface
Clock Speed
(MHz)
Samples per
Clock
Data Width
(Bytes)
System Bus (Data)100483200
DDR Interface1002163200
Hub Interface A6641266
Hub Interface B, C, D66821066
PCI-X133181066
1.3.4System Configurations
Figure 1-1 illustrates an example E7500 chipset-based system configuration for server platforms
using Xeon processors.
Figure 1-1. Example Intel® Xeon™ Processor with 512 KB L2 Cache / Intel® E7500 Chipset
Based System Configuration
ProcessorProcessor
System Memory
200 MHz
DDR
200 MHz
DDR
Hot Plug
Hot Plug
Hot Plug
PCI / PCI-X
PCI / PCI-X
PCI / PCI-X
PCI / PCI-X
PCI / PCI-X
PCI / PCI-X
SMBus
Devices
GPIOs
Super I/O
1–4 FWHs
10/100 LAN
Controller
AC '97
Codec(s)
SMBus 1.1
LPC I/F
AC'97 2.1
Intel
ICH3-S
MCH
16-bit
HI 2.0
16-bit
HI 2.0
®
8-bit
HI 1.5
16-bit
HI 2.0
Intel®
P64H2
P64H2
P64H2
Bandwidth
(MB/s)
USB 1.1, 6 Ports
4 IDE Devices
UltraATA/100
PCI Bus
Agent
PCI
PCI
Slots
Design Guide23
Introduction
This page is intentionally left blank.
24Design Guide
Component Quadrant Layout
Component Quadrant Layout2
The following figures show only general quadrant information, not exact component ball count.
Designers should use only the exact ball assignment to conduct routing analyses. Reference the
following documents for exact ball assignment information.
®
• Intel
• Intel
• Intel
• Intel
Xeon™ Process or wi th 512 KB L2 Cache at 1.80 G Hz, 2 GH z , a nd 2.2 0 G Hz Dat as heet
Platform Stack-Up and Component Placement Overview
Platform Stack-Up and Component
Placement Overview3
3.1Platform Component Placement
Figure 3-1 illustrates the component placement for the Intel Xeon processor with 512 KB L2
cache/Intel E7500 chipset-based customer reference board (E7500 CRB). Table 3-1 lists the
assumptions used for the component placement. Refer to www.ssiforum.org for detailed
information on the SSI (Server System Infrastructure) specification.
T a ble 3-1. Assumptions for System Placement Example
Form Factor (SSI Specification)Number of PCB LayersAssembly
Assumptions
Design Guide31
Platform Stack-Up and Component Placement Overview
3.2Platform Stack-Up
Figure 3-1. Intel
®
E7500 Chipset Customer Reference Board System Placement Example
Figure 3-2 shows the recommended platform stack-up. All layers are 1 oz copper. The processor
requires 2 oz of copper to deliver power and 2 oz of copper to deliver ground. Vias are 10 mil
finished hole with 35 mil anti-pads and 24 mil pads.
Route signal layers as asymmetric st ripline on layers 2 , 4, 5 an d 7. T he sign al layers must ref erence
ground on layer 3 or layer 6 only. Route signals on layers 4 and 5 orthogonally to reduce crosstalk
between the layers.
Intel strongly recommends that system designers use the stack-up shown in Figure 3-2 and
recommendations in Table 3-2 when designing their boards. Intel realizes numerous ways exist to
achieve these targeted impedance tolerances; contact your board vendor for these specifics. Intel
encourages platform designers to perform comprehensive simulation analysis to ensure all timing
specifications are met. This is particularly important if a design deviates from the design guidelines
provided.
32Design Guide
Platform Stack-Up and Component Placement Overview
Figure 3-2. 8 Layer, 50 Ω Board with 5 mil Traces
Layer 1
Dielectric 9.6 mil
Layer 2
Core 5.2 mil
Layer 3
Dielectric 4.3 mil
Layer 4
Core 14.0 mil
Layer 5
Dielectric 4.3 mil
Layer 6
Core 5.2 mil
Layer 7
Dielectric 9.6 mil
Layer 8
Power
Dielectric
SignalSignalSignal
Core
Ground
Dielectric
SignalSignalSignal
Main Core
SignalSignalSignal
Dielectric
Ground
Core
SignalSignalSignal
Dielectric
Power
T able 3-2. E7500 Chipset Customer Reference Board Requirements
Ω single ended, 122 Ω differential pair ± 10% (layers 1 and 8
lower left corner).
• 5 mil trace width and space minimum inner/outer.
• SCSI interface; 6 mil separation within a pair, 20 mil space between adjacent pairs.
• Minimum via size is 0.014 mil finished in a 0.026 mil land with 0.040 mil antipad.
• Approximately 15,000 plated through holes total.
Finish• Solder Mask On Bare Copper (SMOBC)
Soldermask Type• SM-840 minimum web 0.004 mils.
Fabrication• Edge Routed.
Component
Technology
• Through hole / SMT.
• QFP, BGA, Front side.
• Discrete 0603, 0805 Back side.
Design Guide33
Platform Stack-Up and Component Placement Overview
This page is intentionally left blank.
34Design Guide
Platform Clock Routing Guidelines
Platform Clock Routing Guidelines4
To minimize jitter, improve routing, and reduce cost, E7500 chipset-based systems should use a
single chip clock solut ion, the CK 408B. In t his configu ration, t he CK408B provides f our , 1 00 MHz
differential outputs pairs for all of the bus agents, including the ITP connector, and five, 66 MHz
speed clocks that drive all I/O buses. Figure 4-1 shows the implementation of the bus clocks for
this configuration.
For more information on CK408B compliance, refer to the CK408B Clock Synthesizer Specification Specifically for E7500 Chipset DP with ITP System Clock Generator Document.
Table 4-1. CK408B Clock Groups
Clock Group
Name
Host_CLK100 Processor 0, Processor 1, Debug Port and MCH
CLK6666 MCH, ICH3-S, and P64H2
CLK33_ICH3-S33ICH3-S
CLK1414.318ICH3-S and SIO
CLK3333P CI Connec tor, SIO, BMC, and FWH
USBCLK48ICH3-S
Figure 4-1. Intel® E7500 Chipset-Based System Clocking Diagram
CK408B
CPU / CPU# (4)
66BUF (5)
REF0 (1)
USB-48MHz (1)
PCIF (3)
PCI (7)
Host_CLK
CLK66
CLK14
USBCLK
CLK33_ICH3-S
CLK33 x7
CLK66
x3
Processor
Processor
ITP
MCH
Intel® ICH-S
CLK33 (x5)
DDR
Channel A
DIMMclk (x4 pr.)
DDR
Channel B
DIMMclk (x4 pr.)
P
P
P
P
C
C
C
C
I
32 bit
I
I
I
33MHz
D
D
I
D
I
D
M
I
M
I
M
M
M
M
M
M
D
D
I
D
I
D
M
I
M
I
M
M
M
M
M
M
Super I/O
FWH
BMC
Intel® P64H2
PCIclk
x7
P
P
P
P
C
C
C
C
I
I
I
I
Design Guide37
PCIclkx7PCIclk
P
P
P
P
C
C
C
C
I
I
I
I
P64H2P64H2
x7
P
P
P
P
C
C
C
C
I
I
I
I
PCIclkx7PCIclk
P
P
P
P
C
C
C
C
I
I
I
I
P
P
C
C
I
P
P
C
C
I
I
I
PCIclk
x7
x7
P
P
P
P
C
C
C
C
I
I
I
I
Platform Clock Routing Guidelines
4.1Clock Groups
4.1.1HOST_CLK Clock Group
4.1.1.1HOST_CLK Clock Topology
The clock synthesizer provides four sets of 100 MHz differential clock outputs. The 100 MHz
differential clocks are driven to the Processors, the MCH, and the processors’ debug port as shown
in Figure 4-1.
The clock driver differential bus output structure is a “Current Mode Current Steering” output
which develops a clock signal by alternately steering a programmable constant current to the
external termination resistors “Rt.” The resulting amplitude is determined by multiplying IOUT by
the value of Rt. The current IOUT is programmable by a resistor and an internal multiplication
factor so the amplitude of the clock signal can be adjusted for different values of “R”’ to match
impedances or to accommodate future load requirements.
The recommended termination for the differential bus clock is a “Shunt Source Termination.”
Refer to Figure 4-2 for an illustration of this termination scheme. Parallel Rt resistors perform a
dual function, converting the current output of the clock driver to a voltage and matchi ng the driver
output impedance to the transmission line. The series resistors “Rs” provide isolation from the
clock driver's output parasitics, which would otherwise appear in parallel with the termination
resistor Rt.
The value of Rt should be selected to match the characteristic impedance of the motherboard, and
Rs should be between 20 Ω and 33 Ω. Simulations have shown that Rs values above 33 Ω provide
no benefit to signal integrity but only degrade the edge rate.
• Mult0 pin (pin #43) is pulled high – making the multiplication factor 6.
• Iref pin (pin # 42) is connected to ground through a 475 Ω ± 1% resistor – making the Iref
2.32 mA.
Figure 4-2. Source Shunt Termination
L1
L1'
Clock
Driver
LT = L1 + L2 + L4
Rs
Rs
L2
L2'
L3L3'
RtRt
L4
L4'
Processor or
MCH
38Design Guide
Table 4-3. HOST_CLK[1:0]# Routing Guidelines
Layout GuidelineValueIllustrationNotes
HOST_CLK Skew between Agents
Trace Width5 milsFigure 4-4
Differential Pair Spacing20 – 25 milsFigure 4-45,6
Spacing to Other Traces25 milsFigure 4-4
L1, L1’: Clock Driver to Rs
Processor Routing Length –
L2, L2’: Rs to Rs-Rt Node
Processor Routing Length –
L3, L3’: Rs-Rt Node to Rt
Processor Routing Length –
L4, L4’: Rs-Rt Node to Load
MCH Routing Length –
L1, L1’: Clock Driver to Rs
MCH Routing Length –
L2, L2’: Rs to Rs-Rt Node
MCH Routing Length –
L3, L3’: Rs-Rt Node to Rt
MCH Routing Length –
L4, L4’: Rs-Rt Node to Load
Processor to MCH Length Matching (LT)
Processor to Processor Length Matching (LT)
HOST_CLK0 – HOST_CLK1 Length Matching ± 10 mils
Rs Series Termination Value20 – 33
Rt Shunt Termination Value
Platform Clock Routing Guidelines
300 ps total budget:
150 ps for clock driver
150 ps for interconnect
Maintain a minimum S/ h rat i o
of > 5/26
Keep parallel serpentine
sections as short as possible.
Minimize 90 degree bends.
Make 45 degree bends, if
possible.
Ω typical8
Ω ± 10%9
0 – 0.5”Figure 4-213
0 – 0.2”Figure 4-213
0 – 0.2”Figure 4-213
0 – 22”Figure 4-2
0 – 0.5”Figure 4-213
0 – 0.2”Figure 4-213
0 – 0.2”Figure 4-213
0 – 22”Figure 4-2
0.035” ± 0.010”
MCH LT must be 0.076”
longer than Processor LT.
± 10 milsFigure 4-215
Ω ±5%Figure 4-211
Ω ± 1%
49.9
(for 50
Ω board impedance)
Figure 4-2
and
Figure 4-3
Figure 4-4
Figure 4-210
Figure 4-212
1,2,3,4
NOTES:
1. The skew budget includes clock driver output pair to output pair jitter (differential jitter) and skew, clock skew
due to interconnect process variation, and static skew due to layout differences between clocks to all bus
agents.
2. This number does not include clock driver common mode (cycle to cycle) jitter or spread spectrum clocking.
3. The interconnect portion of the total budget for this specification assumes clock pairs are routed on multiple
routing layers and routed no longer than the maximum recommended lengths.
Design Guide39
Platform Clock Routing Guidelines
4. Skew measured at the load between any two-bus agents. Measured at the crossing point.
5. Edge to edge spacing between the two traces of any differential pair. Uniform spacing should be maintained
along the entire length of the trace.
6. Clock traces are routed in a differential configuration. Maintain the minimum recommended spacing between
the two traces of the pair. Do not exceed the maximum trace spacing because this degrades the noise
rejection of the network.
7. Set line width to meet correct motherboard impedance. The line width value provided here is a
recommendation to meet the proper trace impedance based on the recommended stack up.
8. The differential impedance of each clock pair is approximately 2*Zsingle-ended*(1-2*Kb) where Kb is the
backwards cross-talk coefficient. For the recommended trace spacing, Kb is very small, and the effective
differential impedance is approximately equal to 2 times the single-ended impedance of each half of the pair.
9. The single ended impedance of both halves of a differential pair should be targeted to be of equal value. They
should have the same physical construction. If the HOST_CLK traces vary within the tolerances specified,
both traces of a differential pair must vary equally.
10.L ength compensation for the processor socket and package delay is added to chipset routing to match
electrical lengths between the chipset and the processor from the die pad of each. Therefore, the
motherboard trace length for the chipset will be longer than that for the processor.
11.Rs values between 20
12.Rt shunt termination value should match the motherboard impedance.
13.Minimize L1, L2 and L3 lengths. Long lengths on L2 and L3 degrade effectiveness of source termination and
contribute to ringback.
14.The goal of constraining all bus clocks to one physical routing layer is to minimize the impact on skew due to
variations in Er and the impedance variations due to physical tolerances of circuit board material.
15.Length of LT for one processor must match the LT of all other HOST_CLK traces to other processor with
specified tolerance.
Ω and 33 Ωhave been shown to be effective.
Figure 4-3. Clock Skew As Measured from Agent to Agent
BCLK at
Processor
BCLK
at CS
Figure 4-4. Trace Spacing for HOST_CLK Clocks
W
BCLK0BCLK1
h
Ground Plane
BCLK Slew
W
S
S1S1
40Design Guide
4.1.1.2HOST_CLK General Routing Guidelines
When routing the 100 MHz differential clocks, do not split up the two halves of a differential
•
clock pair between layers. Route to all agents on the s ame p hysical rou t ing lay er ref erenced to
ground.
• If a layer transition is required, make sure skew induced by the vias used to transition between
routing layers is compensated in the traces to other agents.
• Do not place vias between adjacent complementary clock traces, and avoid differential vias.
Vias placed in one half of a differential pair must be matched by a via in the other half.
Differential vias can be placed within length L1, between clock driver and Rs, if needed to
shorten length L1.
4.1.1.3CK408 vs. CK408B Requirement
The CK408 and CK408B are pin compatible. The only difference between the two chips is the
CK408B replaces two signals on the CK408 with a fourth HOST_CLK pair for the
In_Target_Probe (ITP) and is preferred by board designers for preliminary testing and validation.
While the CK408B pins need to be connected to the ITP, the CK408 pins require the following
stuffing options:
• Add one 10 kΩ ± 5% pull-up resistor close to the clock driver before the 33 Ω ± 5% (Rs)
(see Figure 4-5) series resistor on each ITP signal trace (CPU3, CPU3#). This would give the
option to use the CK408 instead of the CK408B.
Platform Clock Routing Guidelines
• If deciding to go with CK408, having the 33 Ω± 5% series resistor and 49.9 Ω ± 1% (Rt)
(see
Figure 4-5) parallel resistor is not necessary.
Figure 4-5. Stuffing Options for CK408 and CK408B
VCC3_3
CK408x
CPU3 / S0
VCC3_3
CPU3# /
CPU_STOP#
CK408: 10 kΩ ± 5%
CK408B: no pop
To ITP BCLK1
CK408: no pop
CK408B: Rs
CK408: 10 kΩ ± 5%
CK408B: no pop
To ITP BCLK1
CK408: no pop
CK408B: Rs
Design Guide41
Platform Clock Routing Guidelines
4.1.2CLK66 Clock Group
In the CLK66 clock group, the driver is the clock synthesizer 66 MHz clock output buffer, and the
receiver is the 66 MHz clock input buffer at the MCH, ICH3-S, and P64H2.
Skew Requirements
Clock Driver to MCHX = (3” – 9.5”)
Clock Driver to ICH3-SX = (3” – 9.5”)
Clock Driver to P64H2X – 0.34”
L1L2
R1
)50 Ω± 10%
0
Ω ± 5%
All the clocks in the CLK66 group must have < 100 mil skew
between each other.
1
2
Intel
Intel
MCH,
®
ICH3-S,
®
P64H2
NOTES:
1. For better understanding of the concept, refer to Section 4.1.2.1, Figure 4-7 and Figure 4-8.
2. Assuming no connector.
42Design Guide
4.1.2.1CLK66 Skew Requireme nts
Traces going to the P64H2 could ha ve up to two connectors. Designers should keep in mind that all
T otal Lengths are ref erenced to the MC H length (“X”) and as sume no con nector. Each connector is
equivalent to 0.60 inches of trace. Adding a single connector on the P64H2 trace would reduce the
motherboard trace length by the card length “Z” to X – 0.34” – 0.60” – Z = X – 0.94” – Z (refer
to Figure 4-8). In addition, some OEMs migh t co ns ider having the components o n a ri s er, in which
case the riser card trace length designator “Y” should also be accounted for as yet another factor. In
this case the last equation would become X – 0.34” – 0.60”– Z – 0.60”– Y = X – 1.54”–Y – Z
(refer to Figure 4-9). Note that if a riser is used, the motherboard clock trace must be designed for
the specific riser card trace length and connector.
Figure 4-7. Clock Skew Requirements
43 Ω
CK408B
Platform Clock Routing Guidelines
Total Length = X
MCH
Total Length = X - 0.34"
43 Ω
Resistor must be within
500 mils of CK408B
NOTES:
1. All lengths must be matched within 100 mils of target length.
2. 66 MHz clock lines routed with 25 mils isolation from any other signal.
3. Length from CK408B to MCH must be between 3” and 9.5”.
®
Intel
P64H2
Design Guide43
Platform Clock Routing Guidelines
Figure 4-8. Example of Adding a Single Connector
Total Length = X
43 Ω
CK408B
Motherboard Trace Length
= X - 0.34" - 0.60" - Z
= X - 0.94" - Z
43 Ω
Resistor must be within
500 mils of CK408B
NOTES:
1. All lengths must be matched within 100 mils of target length.
2. 66 MHz clock lines routed with 25 mils isolation from any other signal.
3. Length from CK408B to MCH must be between 3” and 9.5”.
4. Each connector is equivalent to ~ 0.60” of trace.
5. Z is the card trace length.
Figure 4-9. Example of Adding Two Connectors and/or a Riser
Total Length = X
CK408B
43 Ω
MCH
®
Intel
P64H2
Z
MCH
Motherboard Trace Length
= X - 0.34" - 0.60" - Z - 0.60" - Y
= X - 1.54" - Z - Y
Intel
®
P64H2
43 Ω
Resistor must be within 500
mils of CK408B
NOTES:
1. All lengths must be matched within 100 mils of target length.
2. 66 MHz clock lines routed with 25 mils isolation from any other signal.
3. Length from CK408B to MCH must be between 3” and 9.5”.
4. Each connector is equivalent to ~ 0.60" of trace.
5. Each riser is equivalent to ~0.60 + Y where Y is the riser card trace length.
6. The riser must be built with the CLK66 trace length matched to the motherboard routed length.
44Design Guide
Connector
YZ
4.1.3CLK33_ICH3-S Clock
In the CLK33_ICH3-S case, the driver is the clock synthesizer 3 3 MHz clock output buf fer , and the
receiver is the 33 MHz clock input buffer at the ICH3-S.
Figure 4-10. Topology for CLK33_ICH3-S
L1L2
Platform Clock Routing Guidelines
Clock
Driver
T a ble 4-5. CLK33_ICH3-S Routing Guidelines
ParameterRouting Guidelines
Clock GroupCLK33_ICH3-S
TopologyPoint-to-Point
Reference PlaneGround referenced (contiguous over entire length)
Characteristic Trace Impedance (Z
Trace Width5 mils
Trace Spacing25 mils
Trace Length – L10.00” – 0.50”
Trace Length – L23.00” – 9.0”
ResistorR1 = 33
Skew RequirementsMust be matched to ± 100 mils of CLK66
)50 Ω± 10%
0
R1
Intel® ICH3-S
Ω± 5%
Design Guide45
Platform Clock Routing Guidelines
4.1.4CLK33 Clock Group
For the CLK33 clock group, th e driver is the clo ck synthesi zer 33 MHz clock output buffer, and the
receiver is the 33 MHz clock input buffer at the PCI devices on the PCI cards.
Figure 4-11. Topology for CLK33 to PCI Device Down
L1L2
Clock
Driver
.
Table 4-6. CLK33 Routing Guidelines for PCI Device Down
The driver in the CLK14 clock group is the clock synth esizer 1 4.318 MHz clock ou tput bu ff er, and
the receiver is the 14.318 MHz clock input buffer at the ICH3-S, SIO and LPC.
For the USBCLK clock group, the d river is t he clock sy nt hes izer U SB cl ock out pu t buffer, and the
receiver is the USB clock input buffer at the ICH3-S. Note that this clock is asynchronous to any
other clock on the board.
Figure 4-14. Topology for USB_CLK
L1L2
Platform Clock Routing Guidelines
Clock
Driver
T able 4-9. USBCLK Routing Guidelines
ParameterRouting Guideline
Clock GroupUSBCLK
TopologyPoint-to-Point
Reference PlaneGround referenced (contiguous over entire length)
Characteristic Trace Impedance (Z
Trace Width5 mils
Trace Spacing25 mils
Trace Length – L10.00” – 0.50”
Trace Length – L23.00” – 12.00”
ResistorR1 = 33
Skew RequirementsNone – USBCLK is asynchronous to any other clock on the board
Maximum Via Count2
0
R1
Intel® ICH3-S
)50 Ω ± 10%
Ω ± 5%
Design Guide49
Platform Clock Routing Guidelines
4.2Clock Driver Decoupling
The decoupling requirements for a CK408B compliant clock synthesizer are as follows:
• One, 22 µF polarized (decoupling) capacitor placed close to the VDD generation circuitry.
• Eleven, 0.1 µF high-frequency decoupling capacitors placed close to the VDD pins on the
clock drive r.
• Three, 0.1 µF high-frequency decoupling capacitors placed close to the VDDA pins on the
clock drive r.
• One, 10 µF polarized (decoupling) capacitor placed close to the VDDA pins on the clock
driver.
• One, 0.1 µF high-frequency decoupling capacitor placed close to the VDDA generation
circuitry.
• All decoupling capacitors should be placed close to the clock driver pins. Refer to Figure 4-15.
Figure 4-15. Decoupling Capacitors Placement and Connectivity
Ground vias
Ground vias
Ground vias
Ground vias
Ground vias
VSS pins goes through vias on the ground flood to ground plane
Designers must take special care to provide a quiet VDDA supply to the Ref VDD, VDDA and the
48 MHz VDD. These VDDA signals are especially sensitive to switching noise induced by the
other VDDs on the clock chip. They are also sensitive to switching noise generated elsewhere in
the system such as the processor voltage regu lator . It is recomme nded that a ground flood be placed
directly under the clock chip to provide a low impedance connection for the VSS pins. I n addition,
power vias should be distributed evenly throughout the ground flood.
Note:For all power connections to planes, decoupling capacitors, and vias, the maximum trace width
allowable and shortest possible lengths should be used to ensure lowest possible inductance.
4.4EMI Constraints
Clocks are a significant contributor to EMI. The following recommendations can aid in EMI
reduction:
• Maintain uniform spacing between the two halves of differential clocks.
• Route clocks on physical layer adjacent to the VSS reference plane only.
• Turn off all unused clocks.
Platform Clock Routing Guidelines
Design Guide51
Platform Clock Routing Guidelines
This page is intentionally left blank.
52Design Guide
System Bus Routing Guidelines
System Bus Routing Guidelines5
This section covers the system bus source synchronous (data, address, and associated strobes) and
common clock signal routing. Table 5-1 li sts the signals and their corresponding signal types.
Table 5-1. System Bus Signal Groups
Signal GroupTypeSignals
1,2
BPRI#, BR[3:1]#
AGTL+ Common Clock InputSynchronous to BCLK
AGTL+ Common Clock I/OSynchronous to BCLK
AGTL+ Source Synchronous I/O: 4X
Group
AGTL+ Source Synchronous I/O: 2X
Group
AGTL+ StrobesSynchronous to BCLK [1:0]
Async GTL+ Input
Async GTL+ Output
System Bus ClockClockBCLK0, BCLK1
TAP Input
TAP Output
SMBus Interface
Power/OtherPower/Other
1
1
6
6
1
Synchronous to assoc. strobeD[63:0]#, DBI[3:0]#
Synchronous to assoc. strobeA[35:3]#
Asynchronous
Asynchronous
Synchronous to TCKTCK, TDI, TMS, TRST#
Synchronous to TCKTDO
1. These signals do not have on-die termination on the processor. They must be terminated properly on the
motherboard. If the signal is not connected, it must be pulled to the appropriate voltage level through a
1 k
Ω ± 5%resistor.
2. Xeon processors use only BR0# and BR1#.
3. These signals are ‘wired-OR’ signals and may be driven simultaneously by multiple agents. For further details
on how to implement wired-OR signals, refer to the routing guidelines in Section 5.2.1.
4. The value of these pins during the active edge of RESET# determine processor configuration options.
5. SM_VCC has critical power sequencing requirements.
6. Terminations and routing for TAP signals and all debug port signals are found in the
Design Guide
.
ITP700 Debug Port
Design Guide53
System Bus Routing Guidelines
The dual processor topology requires that the MCH be at one end of the bus, Processor 0 be at the
other end of the bus, and Processor 1 be in the middle of the bus (Figure 5-1). The motherboard
routing to Processor 1 must not create a stub on the system bus signals at the socket. This requires
routing into the socket and back out of the socket. For UP operation, the single processor must be
installed in the Processor 0 socket, at the end of the bus. Figure 5-1 shows the recommended dual
processor topology used for system bus routing.
Figure 5-1. Dual Processor System Bus Topology
Processor 0
Motherboard Trace
3.0 – 10.1"3.0 – 10.1 "
Package
Traces
Processor 1
Package Trace
MCH
Refer to Table 5-2 for a summary of the dual processor system bus routing recommendations. Use
this as a quick reference only. The following sections provide more detailed information for each
parameter . Intel strongly reco mmends simulatio n of all signals to ensure the des ign meets setup and
hold times.
54Design Guide
Table 5-2. System Bus Routing Summary
ParameterPlatform Routing Guidelines
Trace Width/Spacing5/15 mils.
2X and 4X Signal Group Line
Lengths
(Agent-to-Agent Length)
DSTBN3[:0]# / DSTBP[3:0]#
and ADSTB[1:0]# Line
Lengths
Common Clock Signal Line
Lengths
T opology Daisy chain with the chipset at one end of the system bus and Processor 0 at
Routing RequirementsNo motherboard contribution to stub length of middle processor (< 35 mil
Reference Plane
Requirements
Motherboard Impedance50
3.0" – 10.1" pin-to-pin.
Total bus length must not exceed 20.2".
Trace lengths must be balanced ± 25 mils with respect to the strobe between
agents to compensate for the stub created by the processor package.
Should follow the same routing rules as the 2X and 4X Signal Group.
A 25 mil spacing should be maintained around each strobe signal.
Do not route differentially.
Common Clock signals should follow the same routing rules as the Data
signals, however no length compensation is necessary.
the other.
End processor must have on-die termination enabled.
trace from via to pad).
All signals within the same strobe group must be routed on same layer for
entire length of bus.
Ground referenced only.
Avoid changing layers when routing system bus signals.
If a layer change must occur, use vias connecting the two reference planes to
provide a low impedance path for the return current. Vias should be as close
as possible to the signal via.
For 2X and 4X signals, ADSTB[1:0]#, and DSTBN3[:0]# / DSTBP[3:0]#:
NEVER ROUTE OVER A PLA N E SPLIT.
Ω ± 10%.
System Bus Routing Guidelines
Design Guide55
System Bus Routing Guidelines
5.1Routing Guidelines for the AGTL+ Source
Synchronous 2X and 4X Groups
The 4X group of signals uses four times the frequency of the base clock, or 400 MHz. The 2X
group uses twice the frequency of the base clock, or 200 MHz. The 2X an d 4X s ignals are listed in
Table 5-3. Table 5-4 lists the 2X and 4X signals with their associated strobes.
Table 5-3. 2X and 4X Signal Groups
2X Group4X Group
A[35:3]#
REQ[4:0]#
Table 5-4. Source Synchronous Signals with the Associat ed Strobes
Routing guidelines for the 2X and 4X signal groups are given in Table 5-2. All 2X and 4X signals
of the same group (refer to Table 5-4) must be routed within ± 25 mils of the same length between
agents and within ± 50 mils of the entire length of the bus.
5.1.1Trace Length Matching
Trace length matching is required within each source synchronous group to compensate for the
package trace length differences between data signals and the associated strobe. This balances the
strobe-to-signal skew in the middle of the setup and hold window. Additional compensation must
be added to account for the capacitive loading effects of the processor socket stubs. Figure 5-2
shows how to implement trace length matching. An example of trace length matching is given in
Example on page 5-57.
Trace length matching consists of matching the pad-to-pad lengths for every signal within a signal
group (e.g., A[35:17]# and ADSTB1#). A pad-to-pad length is measured as follows:
CPU
CPU
Where:
-to-CPU
pad
-to-MCH
pad
CPU
CPU
pin
pin
= CPU0
pad
pad
-to-CPU
-to-MCH
pkg_len
= CPU1
pin
pkg_comp
= Motherboard trace length between Processor 0 and Processor 1.
= Motherboard trace length between Processor 1 and MCH.
pin
pkg_len = Pad to pin length within the package.
0.78 = Compensation due to capacitive loading of processor 1 socket.
CPU1
The package trace lengths for the MCH are available in the Intel® E7500 Chipset Memory
Controller Hub (MCH) Datasheet. The package trace lengths for the Intel Xeon processor with 512 KB L2 cache are available in the matching spreadsheet contained in the Intel
with 512 Cache Signal Integrity Models.
Figure 5-2. Trace Length Matching for the Dual Processor System Bus
®
Xeon™ Processor
CPU
-to-CPU
pin
CPU
pkg_len
Processor 0Process or 1
pin
CPU1
CPU
-to-MCH
pin
pkg_comp
pin
= 0.78 * CPU
MCH
pkg_len
MCH
pkg_len
When length matching, every signal’ s pad to pad length is set equal to each othe r (± 25 mils). This
yields the following equation:
CPU0
CPU0
(Signal 1) + CPU
pkg_len
pkg_len
pad
(Signal 2) + CPU
-to-CPU
-to-CPU
pad
(Signal 1) + 0.78 * CPU1
pad
(Signal 2) + 0.78 * CPU1
pad
pkg_len
pkg_len
(Signal 1) =
(Signal 2)
T o length matc h Signal 1 and Sig nal 2, hold one of the signals con stant, and vary the second sig nal
until the equation is satisfied. Since all the pkg_len values are constant, we can solve for Signal 2:
CPU
-to-CPU
pad
pad
+ 0.78 * CPU1
(Signal 2) = CPU0
(Signal 1) - CPU0
pkg_len
(Signal 1) + CPU
pkg_len
pkg_len
-to-CPU
pad
pad
(Signal 2) - 0.78 * CPU1
(Signal 1)
(Signal 2)
pkg_len
Generally, when length matching a group of signals, a designer will first layout all signals to the
shortest length possible allowed by specification. Then, keeping the longest signal as the constant
value (Signal 1), lengthen all the other signals so that the pad to pad lengths are all equal.
Trace Length Matching Example
Consider two signals, DSTBP0 and HD4, from the same group. Assume a nominal PCB length of
4.00". Calculate CPU to CPU length:
-to-CPU
CPU
pin
CPU
pkg_len
CPU
pkg_len
CPU1
pkg_comp
CPU1
pkg_comp
CPU
-to-CPU
pin
+ CPU1
= 0.350 + 4.000 + 0.273 – 0.150 – 0.117
-to-CPU
CPU
pin
Design Guide57
(HD4) (motherboard trace from Processor 0 to Processor1) = 4.000"
There are five “wired-OR” signals on the system bus. These signals are HIT#, HITM#, MCERR#,
BINIT#, and BNR#. These signals differ from the other system bus signals in that more than one
agent can be driving the signal at the same time. However, Intel recommends that special attention
be given to the routing of these signals in adherence to the layout guidelines presented in Table 5-2.
Timing and signal integrity must be met for the cases where one agent is driving, all agents are
driving, and any combination of agents are driving.
The wired-OR signals should follow the same routing rules as the common clock signals. Intel
recommends that simulations for these signals be performed for each system.
58Design Guide
5.2.2RESET# Topology
Since the processor does not contain on-die termination for the RESET# input signal, these
additional layout guidelines for the RESET# signal are required. The baseboard trace length from
Processor 0's pin to the termination resistor should be 0 to 1 inch.Follow the same routing
guidelines given for common clock signals listed above in this same section.
Figure 5-3. RESET# Topology
VCC_CPU
51 Ω ± 5%
System Bus Routing Guidelines
Processor 1Processor 0
MCH
0.1" - 1. 0"
3.0" - 10.1"3.0" - 10.1"
5.3Routing Guidelines for Asynchronous GTL+ and
Miscellaneous Signals
Table 5-6 enumerates the remainder of the processor signals discussed in this document.
Table 5-6. Asynchronous GTL+ and Miscellaneous Signals (Sheet 1 of 2)
5.3.1Asynchronous GTL+ Signals Driven by the Processor
Follow the topology shown in Figure 5-4 when routing FERR#, IERR#, PROCHOT# and
THERMTRIP#. Note that FERR# is the only signal in this group that connects the processors to the
ICH3-S. IERR#, PROCHOT# and THERMTRIP# connect to oth er mo therb oard logic (su ch as the
Baseboard Management Controller) and may need voltage translation logic, depending on the
motherboard receiver logic devices used. Do not route a stub when routing to the processors.
Figure 5-4. Topology for Asynchronous GTL+ Signals Driven by the Processor
VCC_CPU
Processor 0Processor 1
Intel
ICH3-S
or other logic
0.1" – 3.0"
NOTES:
1. Trace Z0 = 50 Ω.
2. Trace spacing = 10 mil.
0.1" – 10.0"0.1" – 10.0"0.1" – 10.0"
60Design Guide
VCC_CPU
®
56 Ω ± 5% 56 Ω ± 5%
5.3.1.1Proper THERMTRIP# Usage
T o pro tect the pr ocessor s from damag e in over-temperature situations, power to the processor core
must be removed within 0.5 seconds of the assertion of THERMTRIP#. If power is applied to a
processor when no thermal solution is attached, normal leakage currents causes the die temperature
to rapidly rise to levels at which permanent silicon damage is possible. This high temperature
causes THERMTRIP# to go active. Use dual termination on the THERMTRIP# signal. Each
processor’s THERMTRIP# can be routed to its own receiver , or they can be wire-OR’d together. If
routed separately, each signal must be terminated at the receiver end only. All power supply
sources to all processors must be disabled when any installed processor signals THERMTRIP#. In
the reference schematic, the 74AHC74 flip-flop latches the THERMTRIP# signal HIGH after a
PWRGOOD assertion, and LOW after a THERMTRIP# assertion.
Figure 5-5. Recommended THERMTRIP# Circuit
3VSBY
Ω
3904
3904
VCC_CPU
THERMTRIP#
62 Ω
System Bus Routing Guidelines
12 V
74AHC74
10 kΩ1 k
1 kΩ
VCC=3VSBY
SET
DQQ
CLR
100 Ω
THERM_EN to VR
3.3 kΩ
1 kΩ
5.3.2Asynchronous GTL+ Signals Driven by the Chipset
Follow the topology shown in Figure 5-6 when routing A20M#, IGNNE#, INIT#, LINT[1:0],
CPUSLP#, SMI# and STPCLK#. Do not route a stub when routing to the processors.
Figure 5-6. Topology for Asynchronous GTL+ Signals Driven by the Chipset
VCC_CPU
200 Ω ± 5%
NOTES:
1. Trace Z0 = 50 Ω.
2. Trace spacing = 10 mil.
Processor 0Processor 1
0.1" – 3.0"
0.1" – 9.0"0.1" – 9.0"
®
Inte l
ICH3- S
Design Guide61
System Bus Routing Guidelines
5.3.2.1Proper Power Good Usage
Route CPUPWRGD as shown in Figure 5-7. You may choose to isolate PWRGOOD for each
voltage regulator and processor pair in order to recognize individual voltage regulator failures.
Figure 5-7. Topology for PWRGOOD (CPUPWRGOOD)
VCC_CPU
300 Ω ± 5%
NOTES:
1. Trace Z0 = 50 Ω.
2. Trace spacing = 10 mil.
Processor 0Processor 1
0.1" – 3.0"
5.3.2.2Voltage Translation for INIT#
A voltage translator circuit is required for the INIT# signal for all platforms that use the FWH. The
required routing topology for INIT# is given in Figure 5-8. Do not route a stub when ro ut ing to the
processors. Figure 5-9 shows the voltage translator circuit.
Figure 5-8. INIT# Routing Topology
VCC_CPU
200 Ω ± 5%
Processor 0P rocessor 1
0.1" – 9.0"0.1" – 9.0"
ICH 3- S
Inte l
ICH3- S
Intel
®
FW H
®
Voltage
Translator
0.1" – 3.0"
NOTE: T he total trace length between the ICH3-S pin and the Processor 0 pin must be less than 15 inches.
0.1" – 9.0"
0.1" – 9.0"
0.1" – 9.0"
62Design Guide
Figure 5-9. Voltage Translator Circuit
Vcc of Receiver
System Bus Routing Guidelines
300 Ω
± 5%
From Driver
NOTE: T1 and T2 must be referenced to ground.
5.3.3VID[4:0]
Route the VID[4:0] signals of the processor to the VID[4:0] inputs of the voltage regulator
controller. The voltage regulator controller should provide internal pull-up resistors for these
signals. Refer to the VRM 9.1 DC-DC Converter Design Guidelines and the specification of the
voltage controller specific to your design for further details.
Since both processors must operate at the same voltage, the designer should p rovide a way to check
the VID[4:0] signals to ensure a processor does not operate out of specification. (Refer to
Figure 12-3 for more information.
470 Ω
± 5%
470 Ω
± 5%
3904
3904
T1
T1 = 10" max
T2 = 3" max
T2
To Recei ver
5.3.4SMBus Signals
The SMBus signals provide access to the thermal sensor and memory dev ice on the p roces sor. The
signaling protocol used adheres to the specification of the System Management Bus. Refer to
®
Intel
Xeon™ Processor wi th 51 2 KB L2 Cache at 1.8 0 GHz, 2 GHz , and 2.20 G Hz Dat as heet for
details on the Xeon processor implementation and addressing scheme.
Connect the SM_ALER T #, SM_CLK, an d SM_DAT signals to the SMBus contro ller in ad heren ce
to the System Management Bus (SMBus) Specification, Version 1.1. These signals can be connected
to other processors on the sa me SMBus.
The SM_EP_A[2:0] signals set the SMBus address for the memor y dev ice on the processor . These
signals must be set at power up with a unique address per bus. They have an internal 10 kΩ ± 5%
pull-down. To pull the SM_EP_A[2:0] signals to a logic high level, connect each signal to a
100 Ω ± 5% resistor tied to SM_VCC. Refer to the section on SMBus Device Addressing in the
Processor datasheet for addressing details.
Design Guide63
System Bus Routing Guidelines
The SM_TS_A[1:0] signals set the SMBus address for the thermal device on the processor. These
signals must be set at power up with a unique address per bus. The SM_TS_A[1:0] can be set to
logic high, logic low, or a high impedance state giving nine possible combinations of addresses.
Refer to the section on SMBus Device Addressing in the Proces sor datasheet for address ing details.
The SM_TS_A[1:0] signals do not have an internal pull-down and thus must be pulled to VSS or
SM_VCC with a 1 kΩ ± 5% or smaller resistor. Leaving the pins floating achieves a high-Z state.
The SM_WP signal is a write protect signal for the memory device. Pulling this signal to SM_VCC
with a 100 Ω ± 5% resistor enables write protection. SM_WP has an internal 10 kΩ pull-down.
5.3.5System Bus COMP Routing Guidelines
Terminate the processor COMP[1:0] pins to ground through 50 Ω ± 1% resistors. Do not wire the
COMP pins together—connect each pin to its own termination resistor.
Ter minate the MCH HXRCOMP and HYRCOMP with a 25 Ω ± 1% r esi s tor pu l l- down to gro und.
Terminate the MCH HXSWING and HYSWING using a 150 Ω ± 1% res istor pu ll-down to grou nd,
and a 301 Ω ± 1% pull-up to VCC_CPU, respectively. Use two 0.01 µF decoupling capacitors.
5.3.6BR[3:0]# Routing Guidelines
Connect BR[3:0]# as shown in Figure 5-10. The total bus length must be less than 20. 2". BR3# and
BR2# are not used and are pulled to VCC_CPU.
Figure 5-10. BR[3:0]# Connection for DP Configuration
Processor 0, the end processor in a dual processor system, must have its on-die termination
enabled. The termination value must be within 20% of the signal impedance (50 Ω ± 20%). To
enable the on-die termination, pull the ODTEN pin to a high state by terminating it to VCC_CPU
through a 50 Ω ± 20% resistor. Processor 1, the middle agent, must have its on-die termination
disabled. To disable on-die termination, pull the ODTEN pin to a low state by terminating it to
ground through a 50 Ω ± 20% resistor .
64Design Guide
5.3.8TESTHI[6:0] Routing Guidelines
All TESTHI[6:0] pins must be connected to VCC_CPU via pull-up resistors with a termination
value within 20% of the signal imped ance (50 Ω ± 20%). TESTHI[3:0] may all be tied together an d
pulled up to VCC_CPU with a single, 50 Ω ± 20% resistor if desired. TESTHI[6:5] may also be
tied together and pulled up to VCC_CPU with a single 50 Ω ± 20% resistor. However, boundary
scan testing will not be functional if any TESTHI pins are pulled up together. TESTHI4 must
always be pulled up independently from the other TESTHI pins regardless of the usage of
boundary scan.
5.3.9SKTOCC# Signal Routing Guidelines
The SKTOCC# signal is an output fro m the p rocess or u sed as an ind ication o f whether a pr oces so r
is installed or not. It is asserted low when a processor is installed in the socket, and floats when no
processor is present. If this signal is used on the board, the designer can use a pull-up to prevent
floating. SKTOCC# can be used to disable the VRM or VRD output for unpopulated processor
sockets or the power supply output when no processors are installed and other features.
System Bus Routing Guidelines
Design Guide65
System Bus Routing Guidelines
This page is intentionally left blank.
66Design Guide
Memory Interface Routing Guidelines
Memory Interface Routing Guidelines6
The MCH memory interface consists of two DDR memory channels that operate in “lock-step.”
Each channel consists of 64 data and 8 ECC bits. Logically, this is one 144-bit wide memory bus;
electrically, each channel is separate.
This section covers routing guidelines for the DDR interfaces. Note that these guidelines apply to
both channel A and channel B. Each DDR interface has seven signal types: Source Synchronous
Signals, Command Clocks, Source Clocked Signals, Chip Selects, Clock Enable, Receive Enable,
and Miscellaneous. Table 6-1 summarizes the signal groupings. The MCH contains two complete
sets of these signals, one set per channel. Refer to the IntelHub (MCH) Datasheet for details on the signals listed in Table 6-1.
Table 6-1. DDR Channel Signal Groups
GroupSignal
Source Synchronous Signals
Command Clocks
Source Clocked Signals
Chip SelectsCS#[7:0]
Clock EnableCKE
Receive Enable
Miscellaneous
DQS[17:0]
DQ[63:0]
CB[7:0]
CMDCLK[3:0]
CMDCLK[3:0]#
MA[12:0]
RAS#
CAS#
WE#
BA[1:0]
RCVENIN#
RCVENOUT#
DDRCOMP
DDRCVOH
DDRCVOL
DDRVREF[5:0]
®
E7500 Chipset Memory Controller
Design Guide67
Memory Interface Routing Guidelines
6.1DDR Overview
Figure 6-2 and Figure 6-1 show both channels being routed to a single “bank” of eight DIMMs.
The DIMMs are physically interleaved. Intel recommends using this ordering, starting with
Channel B closest to the MCH, for optimal routing.
The platform requires DDR DIMMs to be populated in-order, starting with the 2 DIMMs furthest
from the MCH in a “fill-farthest” approach (see Figure 6-2 and Figure 6-1). This recommendation
is based on the signal integrity requirements of the DDR interface. Intel’s recommendation is to
conduct this check for correct DIMM placement during BIOS initialization. Additionally, it is
strongly recommended t hat al l designs follow the DIMM ordering, SMBus Addressing, Command
Clock routing and Chip Select routing documented in Figure 6-2 and Figure 6-1. This addressing
must be maintained to be compliant with the reference BIOS code supplied by Intel. Designs with
fewer than 3 DIMMs should follow the pattern shown in Figure 6-2 and Figure 6-1.
Figure 6-1. 4 DIMM per Channel Implementation
Fill FourthFill ThirdFill SecondFill First
D
I
MCH
SMBus Address:
Command Clock:
Chip Select:
M
M
B1
04h
0/0#
0/1
Figure 6-2. 3 DIMM per Channel Implementation
Fill ThirdFill SecondFill First
D
I
MCH
M
M
B1
D
I
M
M
A1
00h
0/0#
0/1
D
I
M
M
A1
D
I
M
M
B2
05h
1/1#
2/3
D
I
M
M
B2
D
I
M
M
A2
01h
1/1#
2/3
D
I
M
M
A2
D
I
M
M
B3
06h
2/2#
4/5
D
I
M
M
B3
D
I
M
M
A3
02h
2/2#
4/5
D
I
M
M
A3
D
I
M
M
B4
07h
3/3#
6/7
D
I
M
M
A4
03h
3/3#
6/7
SMBus Address:
Command Clock:
Chip Select:
04h
0/0#
0/1
00h
0/0#
0/1
05h
1/1#
2/3
01h
1/1#
2/3
06h
2/2#
4/5
02h
2/2#
4/5
68Design Guide
Memory Interface Routing Guidelines
The DDR interface requires a nominal impedance (Zo) of 50 Ω ± 10%. Using the recommended
stackup, all routing layers yield 50 Ω nominal impedance when using 5 mil wide traces. Route all
DDR signals 5/15 (5 mils wide with 15 mil spacing) as shown in Figure 6-3 with the exception of
CKE, CMDCLK[3:0], and CMDCLK[3:0]#. For CMDCLK routing rules, refer to Section 6.3 and
Figure 6-8. For CKE routing rules, refer to Section 6.6 and Figure 6-3. Route layers 4 and 5
orthogonal to each other to minimize crosstalk.
Figure 6-3. Trace Width and Spacing for All DDR Signals Except CMDCLK/CMDCLK#
Dielectric 9.6 mil
Core 5.2 mil
Dielectric 4.3 mil
Core 14.0 mil
Dielectric 4.3 mil
Core 5.2 mil
Dielectric 9.6 mil
Layer 1
SignalSignalSignal
Trace
Width
Layer 2
Layer 3
Layer 4
Layer 5
Layer 6
Layer 7
Layer 8
Trace
Spacing
Power
Dielectric
Core
Ground
Dielectric
SignalSignal
Main Core
SignalSignal
Dielectric
Ground
Core
SignalSignal
Dielectric
Power
Trace
Width
Trace
Spacing
Signal
Signal1.4 mil (1 oz)
Signal
Trace
Width
2.1 mil (1 oz + plating)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
2.1 mil (1 oz + plating)
NOTES:
1. Traces on layers 4 and 5 must be routed orthogonally to each other to minimize the eff ects of crosstalk.
2. Source Synch., Source Clocked, and CS# are routed 5/15.
3. CKE is routed 7.5/15.
Design Guide69
Memory Interface Routing Guidelines
6.2Source Synchronous Signal Group
The MCH source synchronous signals are divided into groups consisting of data bits (DQ) and
check bits (CB). An associated strobe (DQS) exists for each DQ and CB group, as shown in
Table 6-2. The MCH supports both x4 and x8 devices, and the number of signals in each d ata group
depends on the type of devices that are populated. For example, if x4 devices are populated, the
72-bit channel is divided into 18 data groups (16 groups con sisting of 4 data b its each, and 2 groups
consisting of 4 check bits each). One DQS is associated with each of these groups (18 total).
Likewise, if x8 devices ar e po pul at ed , the 72 -bi t ch annel is divided into a to tal of ni n e dat a gro ups .
In this case, only 9 of the 18 strobes are used.
In x4 configurations, the high DQS is associated with the high nibble and the low DQS is associated
with the low nibble. In x8 configurations, only the low DQS is used.
Figure 6-4 shows the trace length requirements for the DQ, DQS and CB signals. All signals in a
data group must be length matched to the associated DQSs within ± 100 mils, as shown in
Figure 6-5. In addition, each DQS at a particular DIMM must be length matched to the CMDCLK/
CMDCLK# pair that is routed to that particular DIMM within ± 1.75", as shown in Figure 6-6.
Length matching past the last DIMM connector is not critical. Route all data signals and their
associated strobes on the same layer. Layer changes are only recommended at MCH ball breakout
and at the series resistor. The source synchronous signals require 10 Ω ± 2% series termination
resistors placed close to and before the first DIMM connector, and 22 Ω ± 2% parallel termination
resistors placed as close as possible and after the last DIMM connector (within 0.8").
70Design Guide
Memory Interface Routing Guidelines
Table 6-3. Source Synchronous Signal G roup Routing Guidelines
ParameterIntel® E7500Reference
Signal GroupDQ [63:0], CB[7:0], DQS[17:0]
TopologyDaisy ChainFigure 6-4
Reference PlaneGroundFigure 6-3
MCH to Rtt (Zo)50
MCH to Rtt Trace Width5 milFigure 6-3
Nominal Trace Spacing15 milFigure 6-3
Trace Length – MCH to DIMM11.8” to 6.0”Figure 6-4
Trace Length – Rs to DIMM1< 0.8”Figure 6-4
Trace Length – DIMM to DIMM0.8” to 1.2”Figure 6-4
Trace Length – DIMM to Rtt< 0.8”Figure 6-4
Series Resistor (Rs)10
Termination Resistor (Rtt)22
MCH Breakout Guidelines5/5, < 500 mil
Length Tuning Requirements
Ω ± 10%Figure 6-3
Ω ± 2%Figure 6-4
Ω ± 2%Figure 6-4
DQ to DQS: ± 100 mil
DQS to CMDCLK pair: ± 1750 mil
Figure 6-5
Figure 6-6
Figure 6-4. Source Synchronous Topology
DQ/CB Data Group
Associated DQS
MCH
DQ/CB Data Group
Associated DQS
Intel® MCH to DIMM1
NOTES:
1. Indicated lengths measure from the MCH pin to the DIMM connector pin (including the series resistor).
DDR VTERM
DIMM
to Rtt
(1.25V)
Rtt
Rtt
Rtt
Rtt
Rs
Rs
Rs
Rs
DIMM to
Rs to
DIMM1
DIMM
DIMM to
DIMM
DIMM to
DIMM
DIMMs
Design Guide71
Memory Interface Routing Guidelines
Figure 6-5. Trace Length Matching Requirements for Source Synchronous Routing
Rs
s
l
i
m
0
0
1
+
X
=
)
Q
D
(
a
t
a
D
t
s
e
g
n
o
L
Y
=
)
S
Q
D
(
e
b
o
r
t
S
t
s
e
g
n
o
L
MCH
Shortest Strobe (DQS) = X
DIMM
Rs
Rs
Shortest Data (DQ) = Y – 100 mils
NOTES:
1. The DIMM displayed represents any DIMM. All DIMMs must be length matched within the specified distance.
A simple method to do this is to length match the MCH to the first DIMM within the specified tolerance and
then match all the signals DIMM to DIMM.
2. There are 8 Data lines (DQ) per group. For simplicity purposes, only the longest and the shortest are
represented here.
3. Indicated lengths measure from the MCH die pad to the DIMM connector pin (including the series resistor).
Figure 6-6. DQS To CMDCLK Pair Length Matching
Longest DQS length = x + 1.75"
Shortest DQS length = x - 1.75"
MCH
Rs
DIMM
CMDCLK length = x
CMDCLK# length = x
NOTES:
1. Indicated lengths measure from the MCH die pad to the DIMM connector pin (including the series resistor).
72Design Guide
6.3Command Clock Routing
Only one differential clock pair is routed to each DIMM connector because the MCH only supports
registered DDR DIMMs. All CMDCLK/CMDCLK# termination is on the DIMM modules. Route
each clock and its compliment adjacent to each other. The two complimentary signals (e.g.,
CMDCLK0 and CMDCLK0#) must be length matched to each other within ± 2 mils. Excluding
breakout, the maximum recommended layer changes is one. Ensure that the reference plane does
not change when switching layer.
Table 6-4. Command Clock Pair Routing Guidelines
ParameterIntel® E7500Reference
Signal GroupCMDCLK[3:0], CMDCLK[3:0]#
TopologyPoint to pointFigure 6-7
Reference PlaneGroundFigure 6-8
Differential Trace Impedance (Zo)100
Nominal Trace Width5 milFigure 6-8
Trace Spacing to Complement5 milFigure 6-8
Trace Spacing to Other Traces20 milFigure 6-8
Trace Length – MCH to DIMM12.1” to 10.0”Figure 6-7
Trace Length – MCH to DIMM22.1” to 10.0”Figure 6-7
Trace Length – MCH to DIMM32.1” to 10.0”Figure 6-7
Trace Length – MCH to DIMM42.1” to 10.0”Figure 6-7
MCH Breakout Guidelines5/5, < 500 mil
Length Tuning RequirementsCMDCLK to CMDCLK#: ± 2 mil
Ω ± 10%Figure 6-8
CMDCLK pair to pair: within ± 4.0 in.
CMDCLK pair to DQS pair: ± 1.75 in.
CMDCLK pair to Source Clocked Signal: ± 2.0 in.
Memory Interface Routing Guidelines
Figure 6-7
Figure 6-6
Figure 6-9
Figure 6-7. Command Clock Topology
CMDCLK0 & CMDCLK0#
CMDCLK1 & CMDCLK1#
CMDCLK2 & CMDCLK2#
MCH
NOTES:
1. CMDCLK/CMDCLK# must be matched to within ± 2 mils using package trace length
compensation
2. The total trace length difference of any two CMDCLK/CMDCLK# pairs cannot exceed 4 inches for timing
reasons.
3. For 3-DIMM solutions, treat the CMDCLK3/CMDCLK3# pair as a no connect.
4. Indicated lengths measure from the MCH pin to the DIMM connector pin.
Design Guide73
.
CMDCLK3 & CMDCLK3#
DIMMs
Memory Interface Routing Guidelines
Figure 6-8. Trace Width/Spacing for CMDCLK/CMDCLK# Routing
Dielectric 9.6 mil
Core 5.2 mil
Dielectric 4.3 mil
Core 14.0 mil
Dielectric 4.3 mil
Core 5.2 mil
Dielectric 9.6 mil
Layer 1
CMDCLKCMDCLK#
CMDCLKCMDCLK#
CMDCLK
5 mil
Layer 2
Layer 3
Layer 4
Layer 5
Layer 6
Layer 7
Layer 8
7 mil
Power
Dielectric
Core
Ground
Dielectric
CMDCLK#CMDCLK
Main Core
Dielectric
Ground
Core
CMDCLK#
Dielectric
Power
5 mil20 mil5 mil
Signal
Signal
Signal
Signal
2.1 mil (1 oz + plating)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
1.4 mil (1 oz)
2.1 mil (1 oz + plating)
NOTE: The Intel E7500 MCH may also use a 5 mil space between CMDCLK complements.
Figure 6-9. Length Matching Requirements for Source Clocked Signal, CKE, and CS[7:0]#
Longest CTRL length = x + 2.0"
Shortest CTRL length = x - 2.0"
DIMM
MCH
CMDCLK length = x
NOTES:
1. Indicated lengths measure from the MCH die pad to the DIMM connector pin.
CMDCLK# length = x
74Design Guide
Memory Interface Routing Guidelines
6.4Source Clocked Signal Group Routing
The MCH drives the command clock signals required by the DDR interface; therefor e, no ex ternal
clock driver is required for the DDR interface. The source-clocked signals are “clocked” into the
DIMMs using the command clock signals. Because the MCH drives the command clock signals
and the source-clocked signals together, these signals can be source clocked. That is, the MCH
drives the command clock in the center of the valid window, and the source-clocked signals
propagate with the command clock signal. Therefore, the critical timing is the difference between
the command clock flight time and the source clocked signal flight time. The absolute flight time is
not as critical.
The source-clocked signals have a topology similar to the source synchronous signals. These
signals require parallel termination resistors (Rtt) to DDR VTERM. The MCH requires matching
the lengths of the source-clocked signals to the lengths of the command clocks for each DIMM
within 2.0 inches. For example, if CMDCLK0 and CMDCLK0# are 3 inches long, all source
clocked signals from MCH to the DIMM that CMDCLK0/CMDCLK0# is routed to should be 3
f
Table 6-5. Source Clocked Signal Group Routing Guidelines
inches ± 2.0 inches.
ParameterIntel® E7500Reference
Signal GroupRAS#, CAS#, WE#, MA[12:0], BA[1:0]
TopologyDaisy ChainFigure 6-10
Reference PlaneGroundFigure 6-3
Trace Impedance (Zo)50
Nominal Trace Width5 milFigure 6-3
Nominal Trace Spacing15 milFigure 6-3
Trace Length – MCH to DIMM11.8” to 6.0”Figure 6-10
Trace Length – DIMM to DIMM0.8” to 1.2” Figure 6-10
Trace Length – DIMM to Rtt< 0.8”Figure 6-10
Termination Resistor (Rtt)22
MCH Breakout Guidelines5/5, < 500 mil
Length Tuning RequirementsTo CMDCLK pair: ± 2.0”Figure 6-9
Ω ± 10%Figure 6-3
Ω ± 2%Figure 6-10
Figure 6-10. Source Clocked Signal Topology
DDR VTERM
(1.25V)
RAS#, CAS#, WE#
MA[12:0], BA[1:0]
Rtt
Rtt
MCH
MCH to DIMM1
DIMM to
DIMM
DIMM to
DIMM
DIMMs
NOTE: Indicated lengths measure from the MCH pin to the DIMM connector pin.
Design Guide75
DIMM to
DIMM
DIMM
to Rtt
Memory Interface Routing Guidelines
6.5Chip Select Routing
The MCH provides eight chip select signals. Two chip selects must be routed to each DIMM (one
for each side). The E7500 chip selects for each DIMM must be length matched to the
corresponding clock within ± 2.0 inches and require parallel termination resistors (Rtt) to DDR
VTERM, placed within 3 inches of their associated connector.
Table 6-6. Chip Select Routing Guidelines
ParameterIntel® E7500Reference
Signal GroupCS[7:0]#
TopologyPoint to PointFigure 6-11
Reference PlaneGroundFigure 6-3
Trace Impedance (Zo)50
Nominal Trace Width5 milFigure 6-3
Nominal Trace Spacing15 milFigure 6-3
Trace Length – MCH to DIMM11.8” to 9.6”Figure 6-11
Trace Length – MCH to DIMM21.8” to 9.6”Figure 6-11
Trace Length – MCH to DIMM31.8” to 9.6”Figure 6-11
Trace Length – MCH to DIMM41.8” to 9.6”Figure 6-11
Trace Length – DIMM to Rtt< 3.0”Figure 6-11
Termination Resistor (Rtt)22
MCH Breakout Guidelines5/5, < 500 mil
Length Tuning RequirementsTo CMDCLK pair: ± 2.0”Figure 6-9
Ω ± 10%Figure 6-3
Ω ± 2%Figure 6-11
Figure 6-11. Chip Select Topology
CS0#
CS1#
CS2#
CS3#
MCH
NOTES:
1. For 3-DIMM solutions, treat CS6# and CS7# as a no connect.
2. Indicated lengths measure from the MCH pin to the DIMM connector pin, and from the DIMM connector to
the parallel termination resistor pin.
CS4#
CS5#
CS6#
CS7#
MCH to DIMM
DDR VTERM (1.25V)
DIMMs
Rtt
DIMM
to Rtt
76Design Guide
6.6Clock Enable Routing
The MCH provides a single clock enable (CKE) signal. This signal is used during initialization to
indicate that valid power and clocks are being applied to the DIMMs. Because the CKE signal has
higher loading, it requires a lower impedance. The recommended impedance fo r the CKE signal is
40 Ω. This can be achieved using a 7.5 mil wide trace on the recommended stackup (refer to
Figure 6-3). It is acceptable to route the CKE signal 5 mils wide when breaking out of the MCH.
The CKE signal requires a parallel termination resistor (Rtt) to DDR VTERM placed as close to
the last DIMM connector as possible.
Ta ble 6-7. Clock Enable Routing Guidelines
ParameterIntel® E7500Reference
Signal GroupCKE
TopologyDaisy Chain with StubsFigure 6-12
Reference PlaneGroundFigure 6-3
Trace Impedance (Zo)40
Nominal Trace Width7.5 milFigure 6-3
Nominal Trace Spacing15 milFigure6-3
Trace Length – MCH to DIMM11.8” to 6.0”Figure 6-12
Trace Length – DIMM to DIMM0. 8” to 1.2” Figure 6-12
Trace Length – CKE Stub< 300 milFigure 6-12
Trace Length – DIMM to Rtt< 0.8”Figure 6-12
Termination Resistor (Rtt)22
MCH Breakout Guidelines5/5, < 500 mil
Length Tuning RequirementsTo CMDCLK pair: ± 2.0”Figure 6-9
Memory Interface Routing Guidelines
Ω ± 10%Figure 6-3
Ω ± 2%Figure 6-12
Figure 6-12. CKE Topology
DDR VTERM
(1.25V)
MCH
CKE
NOTE: Indicated lengths measure from the MCH pin to the DIMM connector pin.
Design Guide77
CKE Stub
MCH to DIMM1
DIMM to
DIMM
DIMM to
DIMMs
DIMM
DIMM to
DIMM
DIMM
to Rtt
Rtt
Memory Interface Routing Guidelines
6.7Enable Signal (RCVEN#)
The MCH uses the “receive enable” (RCVEN#) signal to determine the approximate DIMM
round-trip flight time (command flight + data flight). Two pins exist on the MCH to facilitate the
use of RCVEN#: RCVENOUT# and RCVENIN#. RCVENOUT# is an output of the MCH, and
RCVENIN# is an input to the MCH. The board designer must connect RCVENOUT# to
RCVENIN#. The length of the RCVEN# signal trace must be 15" ± 100 mils. Figure 6-13
illustrates the routing recommendations of the RCVEN# signal.
Figure 6-13. Receive Enable Signal Routing Guidelines
MCH
RCVENOUT#
RCVENIN#
DDR VTERM
(1.25V)
47
RCVEN# Total Length:
15 inches ± 100 mils
Ω ± 2%
78Design Guide
6.8Miscellaneous Signals
The MCH uses a compensation signal to adjust buf fer characteristics and output vo ltage swing over
temperature, process, and voltage skew. Calibration is done periodically by sampling the
DDRCOMP, DDRCVOH, and DDRCVOL pins on the MCH. Connect DDRCOMP to the DDR
termination voltage (1.25 V) through a 6.81 Ω ± 1% resistor as illustrated in Figure 6-14, and place
the resistor within 1 in. of the MCH. Likewise, keep the voltage divider networks within 1 in. of the
MCH (see Figure 6-15).
Figure 6-14. DDRCOMP Resistive Compensation
Memory Interface Routing Guidelines
DDR VTERM
(1.25 V )
MCH
DDRCOMP
6.81 Ω ± 1%
<1"
Figure 6-15. DDRCVOL and DDRCVOH Resistive Compensation
DDR VDD
(2.5V)
MCH
DDRCVOHDDRCVOL
< 1"
7 kΩ ±
1%
13 kΩ ± 1%
DDR VTERM
(1.25V)
0.01 µF
0.01 µF
MCH
< 1"
13 kΩ ±
1%
DDR VTERM
(1.25V)
7 kΩ ± 1%
0.01 µF
0.01 µF
Design Guide79
Memory Interface Routing Guidelines
6.9DDR Reference Voltage
The DDR system memory reference voltage (VREF) is used by the DRAM devices and the MCH
to determine the logic level being driven on the data, strobe, and control signals. VREF of the
receiving device must track changes in VTT to maximize DDR interface margin. If a voltage
regulator is used, it must reference VTT (See Figure 6-16). If a local resistor divider is used, VREF
and VTT must have a common source voltage between them (i.e., both VREF and VTT are deriv ed
from the same voltage plane), and 1% res ist or s s hou l d be u sed (See Figure 6-17). Decouple VREF
locally at the divider and DIMMs/MCH using one 0.1uF capacitor per VREF pin.
Figure 6-16. DDR VREF Voltage Regulator
DDR VTT
(1.25V)
DDR VDD
(2.5V)
Ref
Voltage
Regulator
Vin
Vout
DDR VREF
(1.25V)
0.1 µF
Figure 6-17. DDR VREF Voltage Divider
DDR VREF
(1.25V)
0.1 µF
1%
1%
DDR VDD
(2.5V)
Vin
Voltage
Regulator
Vout
DDR VTT
(1.25V)
80Design Guide
6.10DDR Signal Termination
Place a 1.25 V termination plane on the top layer just beyond the DIMM connector furthest from
the MCH on each channel, as shown in Figure 6-18. The VTERM island must be at least 50 mils
wide. Use this termination plane to terminate all DIMM signals, using one 22 Ω ± 2% resistor per
signal. Decouple the VTERM plane using one 0.1 µF decoupling capacitor per two termination
resistors. In addition, place one 100 µF Tantalum capacitor on each end of each termination island
for bulk decoupling. Each decoupling capacitor must have at least 2 vias between the top layer
ground fill, and the internal ground plane. Refer to Figure 6-18.
Figure 6-18. DDR VTerm Plane
Memory Interface Routing Guidelines
Two Vias Per 1 Capacitor
1.25V Vterm Fill
One Rtt per signal
1.25V
Vterm F ill
Ground Fill
on Top Layer
Two V ia s P e r 1
Capacito r to th e
Internal Gr ou n d
Plane
to the Internal Gro und
Plane
DIMM8 (Fu rthe st from MCH)
50 mils
minimum
Ground Fill on
Top Layer
DIMM7
DIMM6
One 0.1 µF De coup ling
Capacitor per 2 Termination
Resistors or (2 Caps/Rpack)
50 mils
minimum
DIMM to Rtt
(0.8" max)
One 100 µF Tantalum
Capacito r at E a ch E n d
of Each Isla nd
One 0.1 µF
decoupling
capactior per 2
termination
resistors
Design Guide81
Memory Interface Routing Guidelines
6.11Decoupling Requirements
Decouple the DIMM connectors as shown in Figure 6-19. Place six ceramic 0.1 µF (0603)
capacitors between adjacent DIMM connectors. Place ten Tantalum 100 µF capacitors per channel
around the DIMM connectors, keeping them within 0.5" of the edge of the DIMM connectors.
Again, be sure to implement two vias per capacitor (ceramic and tantalum) to the internal ground
plane.
Figure 6-19. DIMM Decoupling
10 Tantulum 10 0 µF
Capacitors/Channel
Around DIMMs
DIMM
2 Vias Per Capa citor to
Internal Ground Plane
DIMM
DIMM
DIMM
6 Ceramic 0.10 µF Caps
(0603) B etween DIMM
Pairs
82Design Guide
Hub Interface
Hub Interface7
7.1Signal Naming Convention
Figure 7-1 has the Hub Interface 2.0 and Hub Interface 1.5 signal naming convention for each
component. This figure is intended to give a quick naming cross reference to designers. The
specific guidelines and implementations on these signals are given in the following sections. Note
that throughout the document, the ‘x’ part of the MCH signal has been dropped for simplicity.
Figure 7-1. Signal Naming Convention on Both Sides of the Hub Interfaces
®
Intel
P64H2
PUSTRBS
HI2.0
PUSTRBS_x
MCH
PUSTRBF
PSTRBS
PSTRBF
HI_[#]
HI_VSWING
HI_VREF
HI_RCOMP
®
Intel
ICH3-S
HI_STBS
HI_STBF
HI[#]
HITERM
HIREF
PUSTRBF_x
PSTRBS_x
PSTRBF_x
HI_x[#]
1
1
2
HI1.5
1
1
HISWNG_x
HIVREF_x
HIRCOMP_x
HI_STBS
HI_STBF
HI_A[#]
HISWNG_A
HIVREF_A
1
1
2
1
1
2
HICOMP
NOTES:
1. These signals have individual resistor dividers. For specific values, refer to Figure 7-5 and Figure 7-8.
2. These signals have individual pull-up resistors. For specific values, refer to Figure 7-6 and Figure 7-9.
3. Signal names for HI2.0 on the MCH: x = B, C, or D.
HIRCOMP_A
2
Design Guide83
Hub Interface
7.2Hub Interface 2.0 Implementation
The MCH, and P64H2 ballout assignments are optimized to simplify the hub interface routing
between these devices. To allow for greater flexibility in design, a connector can be placed on the
interface to access a HI2.0 agent that resides on an adapter card. The typical card implementation
uses an extension to the 3.3 V PCI-64 connector that provides an additional 70 pins for HI2.0.
Power, JTAG and SMBus signals are taken from the PCI portion of the connector. The remaining
PCI signals are unused. This approach provides the flexibility to allow either a
PCI/PCI -X card or a HI2.0 card, to be populated in the slot.
For the 16-bit Hub Interface, HI[7:0] and HI[20] are associated with PSTRBF and PSTRBS, and
HI[15:8] and HI[21] are associated with PUSTRBF and PUSTRBS. HI[19:16] are common clock
signals; they are sampled using CLK66. The three hub interfaces on the MCH are functio nally and
electrically identical. Therefore, these guidelines apply to all three hub interfaces.
Table 7-1. Hub Interface 2.0 Signal/Strobe Association
This section documents the routing guidelines for the Hub Interface 2.0. The Hub Interface 2.0
signal groups are listed in Table 7-2. The general routing guidelines for the Hub Interface 2.0
signals are given in Table 7-3.
The Hub Interface signals must be routed directly from the MCH to P64H2 with all signals
referenced to ground. Maintain a consistent ground reference plane at all times. In addition, route
all signals within a data group (consisting of nine bits of data and a pair of strobes) on the same
layer and reference them to the same ground plane. Keep layer transitions to a minimum. If a layer
change is required, use only two vias per net and keep all signals within a data group on the same
layer.
Route the Hub Interface 2.0 data signal traces 5 mils wide using the recommended stackup. There
must be 15 mils spacing between signal traces (5/15). Each strobe signal must have a minimum of
35 mils of spacing from any adjacent signals to minimize effects that cause signal degradation. To
break out of the MCH and P64H2 package, the hub interface data signals can be routed 5/5. The
signals must separate to 5/15 (or strobes to 5/35) within 0.5 inch of the package.
Hub Interface 2.0 requires package length compensation, which is similar to the system bus
package length compensation. For E7500 chipset component package lengths, refer to the
component datasheets.
For Hub Interface 2.0 devices on the motherboard, package trace length matching of ± 0.25 inch
(including package length compensation) is required among all signals within a data group. If the
hub device is on an adapter, length matching of ± 0.125 inch (including package length
compensation) is required among all signals within a data group. The hub interface strobe trace
lengths must be 0 to 1.0 inch shorter than the longest hub interface data trace.
Figure 7-2 depicts the length matching rules for a hub device on the motherboard. All of the Hub
Interface Data signals must be length matched within 0.25 inch. The figure shows HI[x] and HI[y]
with the maximum allowed difference in length, while HI[z] is somewhere in the middle. The
strobes in each strobe pair (PSTRBF and PSTRBS; PUSTRBF an d PUSTRBS) are also matched
within 0.25 inch. However, the absolute length of the strobe pair is adjusted according to the
longest Hub Interface Data line. The upper pair shows the case where one of the strobes is the
same exact length as the longest Hub Interface Data line (which is the longest possible length one
of the strobes can be). In this case, the other strobe must be equal to or shorter than it, but by no
more than 0.25 inch. The lower strobe pair shows the case where one of the strobes is exactly
1.0 inch shorter than the longest Hub Interface Data line (which is the shortest possible length one
of the strobes can be). In this case, the other strobe must be equal to or longer than it, but by no
more than 0.25 inch.
Design Guide85
Hub Interface
Figure 7-2. Hub Interface 2.0 Length matching
HI[x]
HI[y]
HI[z]
PSTRBF
PSTRBS
- OR -
PSTRBF
PSTRBS
NOTES:
1. All signal lines with arrows depict the total length of the signal including the mother board trace length, MCH
package trace length, and Hub Interface 2.0 device trace length.
2. PUSTRBF and PUSTRBS length matching is the same as for PSTRBF and PSTRBS.
3. This figure is only an example for an implementation with the device on the motherboard. For an
implementation with the hub interface device on a riser card, simply replace both instances of 0.25”
with 0.125”.
4. In the example above, HI[x], HI[y], and HI[z] represent Hub Interface data signals. The other six data signals
in the group must also be matched within 0.25”. The associated strobe pair must be within 1.0” of the longest
data signal.
0.25"
0.25"
1.0 "
Hub Interface 2.0 has a minimum trace length requirement of 3 inches, and a maximum trace
length requirement of 20 inches for a device on the motherboard implementation for all hub
interface signals (using an internal routing layer on the recommended stackup). However, for a
device on an adapter card plugged in a hub interface 2.0 connector, the maximum motherboard
trace length is 14 inches. For a riser card topology, the maximum trace length would reduce to
3 inches to (11-Y) inches, where Y is the riser card trace length. The riser must be built to not
exceed the maximum trace length with the motherboard routed length.
Figure 7-3. Hub Interface 2.0 Routing Guidelines for Device Down Solutions
NOTE: The 14 inch maximum length allows for a single connector and 3 inch adaptor card trace length. The
PCI connector is an equivalent 3 inch electrical length. The maximum motherboard trace length must be
shortened if additional trace is allocated for the trace of a riser card, making sure to also subtract the
additional equivalent trace of a second connector.
7.2.2Hub Interface 2.0 Generation/Distribution of Reference
Voltages
The nominal Hub Interface 2.0 reference voltage is 0.350 V ± 5%. Each Hub Interface 2.0 on the
MCH has a dedicated HIVREF pin to sample this reference voltage. Similarly, the P64H2 has a
dedicated reference voltage pin. In addition to the reference voltage, a reference swing voltage
must be supplied to control buffer voltage swing characteristics. The nominal Hub Interface 2.0
reference swing voltage should be 0.8 V ± 5% for the MCH and P64H2. Each Hub Interface 2.0 o n
the MCH has a dedicated HISWNG pin to sample this reference swing voltage. The P64H2 has a
dedicated reference swing voltage pin as well. Both of these reference voltages can be generated
locally with a single voltage divider circuit. Figure 7-5 shows an example voltage divider circuit.
T able 7-4. Hub Interface 2.0 Reference Circuit Specifications
Reference Voltage
Specification (V)
0.350 ± 5%
Reference Swing Voltage
Specification (V)
For P64H2 = 0.8 ± 5%
For MCH = 0.8 ± 5%
1.2 V Voltage DIvider
Resistor Values (
Circuit
Recommended
R1 = 392 ± 1%
R2 = 499 ± 1%
R3 = 453 ± 1%
1.8 V Voltage DIvider
Ω)
Recommended
Resistor Values (Ω)
R4 = 261 ± 1%
R5 = 332 ± 1%
R6 = 750 ± 1%
Circuit
Design Guide87
Hub Interface
Figure 7-5. Hub Interface 2.0 with Locally Generated Voltage Divider Circuit
MCH
HISWNG_x
HIVREF_x
C2
C2
1.2 V
R3
R2
R1
C1
C1
0.8 V
C1
0.35 V
0.8 V
R6
R5
R4
1.8 V
Intel® P64H2
HI_VSWING
C2C1
HI_VREF
C2
0.35 V
The resistor values R1, R2, R3, R4, R5, and R6 must be rated at ± 1% tolerance. The selected
resistor values must also ensure that the reference voltage and reference swing voltage tolerance
are maintained over the input leakage specification. A 0.1 µF capacitor (C1 in the above circuits)
should be placed close to each resistor divider, and a 0.01 µF bypass capacitor (C2 in the above
circuits) should be placed near each reference voltage pin. If the length of the trace from the
voltage divider to the pin is greater than 1", place more than one 0.01 µF capacitor near the
reference voltage pin. The trace length from the voltage divider circuit to the corresponding pin
must be no longer than 3.5 inch es.
Both the voltage reference and voltage swing reference signals should be r outed 20 mils to 25 mils
from all other signals.
7.2.3Hub Interface 2.0 Resistive Compensation
The hub interface uses a resistive compensation signal (HIRCOMP_x) to compensate buffer
characteristics across temperature, voltage, and process. The HIRCOMP_x resistor values are
given in Table 7-5. Figure 7-6 shows the RCOMP_x circuits .
ComponentTrace ImpedanceRCOMP Resistor Value RCOMP Resistor Tied To
MCH50 Ω ± 10%R1 = 24.9 Ω ± 1%VCC1.2
®
P64H250 Ω ± 10%R2 = 61.9 Ω ± 1%VCC1.8
Intel
Figure 7-6. Hub Interface 2.0 RCOMP Circuits
1.2 V
®
MCH
R1
HIRCOMP_x
Intel
P64H2
HI_RCOMP
1.8 V
R2
88Design Guide
7.2.4Hub Interface 2.0 Decoupling Guidelines
To improve I/O power delivery, use two 0.1 µF capacitors per component (i.e., MCH, P64H2).
These capacitors should be placed within 150 mils of each package, adjacent to the rows that
contain the hub interface. If the layout allows, wide metal fingers running on the VSS side of the
board should connect the VCC1.8/VCC1.2 side of the capacitors to the VCC1.8/VCC1.2 power
pins. Similarly, if layout allows, metal fingers running on the VCC1.8/VCC1.2 side of the board
should connect the ground side of the capacitors to the VSS power pins.
7.2.5Unused Hub Interface 2.0 Interfaces
Terminate unused Hub Interface 2.0 interfaces as described below:
• All hub interface data and strobe signals can be left as no connects.
• HIVREF and HISWNG must remain connected to the reference voltage divider circuits (refer
to Figure 7-5). RCOMP must be pulled up to 1.2 V.
7.3Hub Interface 1.5 Implementation
Hub Interface
The Hub Interface 1.5 signals HI[7:0] are associated with HI_STBS/HI_STBF. For those familiar
with the Hub Interface 1.0 mode, HI_STBF and HI_STBS are called HI_STB# and HI_STB,
respectively.
Figure 7-7. 8-Bit Hub Interface 1.5 Routing
HI_STBF
®
Intel
ICH3-S
This section documents the routing guidelines for the Hub Interface 1.5 that is responsible for
connecting the MCH to the ICH3-S. Hub Interface 1.5 supports parallel termination mode only,
therefore the DPRSLPVR pin on the ICH3-S must be left as No Connect (NC); this signal has an
internal pull-down.
The MCH and ICH3-S ball assignments are optimized to simplify the hub interface routing
between these devices. Route the hub interface signals directly from the MCH to ICH3-S with all
signals referenced to ground. Keep layer transitions to a minimum. If a layer change is required,
use only two vias per net, and keep all data signals and associated strobe signals on the same layer.
The Hub Interface 1.5 signal groups are listed in Table 7-6. The general routing guidelines for the
Hub Interface 1.5 signals are given in Table 7-7.
Using the recommended stackup, the Hub Interface 1.5 data signal traces must be routed 5 mils
wide. There must be 15 mils spacing between traces (5/15). To break out of the MCH and ICH3-S
packages, the Hub Interface data signals can be routed 5/5. The signals must be separated to 5/15
within 0.3 inch of the package.
For Hub Interface 1.5 devices on the motherboard, each strobe signal trace must be the same
length, and each data signal trace must be matched within ± 0.1 inch.
7.3.2Hub Interface 1.5 Generation/Distribution of Reference
Voltages
The nominal Hub Interface 1.5 reference voltage is 0.35 V ± 5%. The 8-bit Hub Interface on the
MCH has a dedicated HIVREF pin to sample this reference voltage. In addition to the reference
voltage, a reference swing volt age must be s uppl ied to contr ol buf fer voltage s wing cha racteristi cs.
The nominal Hub Interface 1.5 reference voltage swing must be 0.8 V ±
0.7 V ±
5% for the ICH3-S. This vol ta g e is sampled by the MCH using HISWING, and is samp led
by the ICH3-S using HITERM. (see Table 7-8). Both HISWNG and HITERM can be generated
locally with a single voltage divider circuit as shown in Figure 7-8.
T a ble 7-8. Hub Interface 1.5 Reference Circuit Specifications
The values of R1, R2, R3, R4 and R5 must be rated at ± 1% tolerance. The selected resistor values
must also ensure that the reference voltage and reference swing voltage tolerance are maintained
over the input leakage specification. A 0.1 µF capacitor (C1 in Figure 7-8) should be placed within
0.5 inch of each resistor divider, and a 0.01 µF bypass capacitor (C2 in Figure 7-8) should be
placed within 0.25 inch of reference voltage pins. If the length of the trace from the voltage divider
to the pin is greater than 1 inch, place more than one 0.01 µF capacitor near the reference voltage
pin. The trace length from the voltage divider circuit to the HIREF and HUBREF pins must be no
longer than 3.5 inches.
Both the voltage reference and voltage swing reference signals should be routed at least 20 mils to
25 mils from all other signals.
7.3.3Hub Interface 1.5 Resistive Compensation
HITERM
Intel
ICH3-S
HIREF
®
The hub interface uses a resistive compensation signal (RCOMP) to compensate buffer
characteristics for temperature, voltage, and process. The HIRCOMP resistor values are given in
Table 7-9. Figure 7-7 shows the RCOMP_x circuits.
T a ble 7-9. Hub Interface 1.5 RCOMP Resistor Values
ComponentTrace ImpedanceRCOMP Resistor V alue RCOMP Resistor Tied To
To improve I/O power delivery, use two 0.1 µF capacitors per each component (i.e., the ICH3-S
and MCH). These capacitors should be placed within 150 mils of each package, adjacent to the
rows that contain the hub interface. If the layout allows, wide metal fingers running on the VSS
side of the board should connect the VCC_1.8/VCC_1.2 side of the capacitors to the VCC_1.8/
VCC_1.2 power pins. Similarly, if layout allows, metal fingers running on the VCC_1.8/VCC_1.2
side of the board should connect the ground side of the capacitors to the VSS power pins.
92Design Guide
Intel® 82870P2 (P64H2)
Intel
8.1PCI/PCI-X Design Guidelines
Table 8-1. PCI/PCI-X Frequencies
82870P2 (P64H2)8
The 82870P2 (P64H2) is a peripheral chip that performs PCI/P CI- X bridg ing functions between
Hub Interface and the PCI bus. The P64H2 is an integral part of the E7500 chipset, bridging the
MCH and the PCI/PCI-X bus. On the primary bus, the P64H2 utilizes a 16-bit data bus to interface
with the Hub Interface 2.0, and on the secondary bus, it supports two 64-bit PCI bus segments.
Either of the secondary PCI/PCI-X bus interfaces can be configured to operate in PCI or PCI-X
mode. Each PCI/PCI-X interface contains an I/OxAPIC with 24 interrupts and a hot plug controller
that supports each PCI/PCI-X bus segment.
The P64H2 contains two PCI/PCI-X Interfaces. Th e PCI Interf ace has a 33/66 MHz bus s peed, and
the PCI-X interface has a 66/100/133 MHz bus speed (see Table 8-1).
PCI
FrequencyMaximum SlotsVoltage
33 MHz63.3 V , 5 V
66 MHz23.3 V
®
PCI-X
FrequencyMaximum SlotsVoltage
66 MHz43.3 V
100 MHz23.3 V
133 MHz13.3 V
NOTE: Frequencies specified are not the only ones supported, rather the maximum allowed in the
configuration.
Intel simulated the PCI/PCI-X bus topologies shown in Section 8.1.1 and Section 8.1.2. If a
platform implements a PCI/PCI-X topology not found in the following sections, it is the
responsibility of the system designer to ensure the system meets the specified timings. The
recommended lengths specified are not intended to replace thorough system simulations and
validation.
Design Guide93
Intel® 82870P2 (P64H2)
8.1.1PCI/PCI-X Routing Requirements (No Hot Plug)
The P64H2 supports a lar ge number of PCI/PC I-X confi guratio ns. The basic topo logy of the bus is
shown in Figure 8-1. Multiple slots are connected in a daisy chain topology with the device(s)
down on the motherboard at the end of the daisy chain. Table 8-2 documents the lengths for the
configurations Intel simulated.
NOTE: During simulation, slot to slot lengths were held constant for some configurations. Therefore, no range
Inte l
®
P64H2 to Slot
Slot 1
P64H2
Slot
to Slot
®
P64H2 PCI/PCI-X Configuration Length Requirements
®
P64H2
Configuration
can be given for these length requirements.
Intel
to Slot
Slot 2
Device Down
Slot to SlotSlot to Device Down
Slot to
Device
Down
94Design Guide
8.1.2PCI/PCI-X Hot Plug Routing Requirements
The P64H2 supports a large number of PCI/PCI-X Hot Plug configurations. The Hot Plug topology
of the bus is shown in Figure 8-2. Hot Plug switches are connected in a daisy chain topology with
the device(s) down on the motherboard at the end of the daisy chain. Table 8-3 documents the
lengths for the configurations that Intel simulated.
NOTE: During simulation, slot to slot lengths were held constant for some configurations. Therefore, no range
Intel
®
P64H2 to
Switch
Switch1
Switch to
Slot
P64H2
Switch to
Switch
®
P64H2 Hot Plug Configuration Length Requirements
®
P64H2 to
Configuration
can be given for these length requirements.
Intel
Switch
Switch2
Device Down
Switch to SlotSwitch to Switch
Switch to
Switch to
Slot 1
Slot 2
Slot
Device
Down
Switch to
Device Down
Design Guide95
Intel® 82870P2 (P64H2)
8.1.3Clock Configuration
All PCI clocks must be disabled in the BIOS for any unused/unpopulated PCI/PCI-X slots. The
PxPCLKO[5:0] pins can each be disabled by writing to the Disable PCLKOUT 5 – 0 bits (DPCLK,
bits 15:10, config register offset 40h in each bridge). These clocks function the same in Serial and
2-Slot Parallel modes. In 1-Slot Serial Mode, the PxPCLKO[5:0] signals are all driven low when
the clock to the slot is disabled by the hot plug controller, regardless of the DPCLK bits. Once the
Hot Plug controller connects the clock to the slot, these clocks are enabled ag ain—which clocks are
enabled does depend on DPCLK at this point. It is expected that PxPCLK[0] will be connected to
the PCI slot in Single Slot Parallel Mode.
Figure 8-3. Hot Plug Clock Configuration
Inte l
®
L1
33 ΩL3
P64H2
Table 8-4. Hot Plug Clock Routing Length Parameters
1. The clock signal and feedback loops are closely related. Refer to Figure 8-4 for L2, and Figure 8-5 for L
≤ 1.0L
fbi
fbi
– 2.5
– 2.5
1
1
96Design Guide
1
L
fbi
1
L
fbi
.
fbi
8.1.4Loop Clock Configuration
You must tie PxPCLKO[6] to PxPCLKI because this clock always runs and is needed by the
internal PCI PLLs to properly align output signals with the external clocks by removing clock
insertion delay. The PxPCLKO[6] signal does not have to be routed through a bus switch before
returning to PxPCLKI.
33 MHz / No HP3.5 – 5.52.9 – 7.9
66 MHz / No HP4.5 – 5.53.9 – 4.9
66 MHz / With HP0.25 – 1.07.0 – 12.0
100 MHz / No HP
100 MHz / With HP4.5 – 5.53.9 – 4.9
133 MHz / No HP0.25 – 1.0L2 + 2.5
133 MHz / With HP3.5 – 4.05.5 – 5.7
NOTES:
1. The clock signal and feedback loops are closely related. Refer to Figure 8-4 for L2 and Figure 8-5 for L
(inches)L
fbo
≤ 1.0L2 + 2.5
fbi
(inches)
1
1
.
fbi
Design Guide97
Intel® 82870P2 (P64H2)
8.1.5IDSEL Implementation
Designers should use a 100 Ω series coupling resistor on the IDSEL signal when implementing
PCI-X. Though the PCI-X Addendum PCI Local Bus Specification, Revision 1.0 calls for a 2 kΩ
resistor, the current specification, PCI-X Addendum to the PCI Local Bus Specification, Revision
1.0a allows for other resistor values. See Figure 8-6 for an example of how to implement the
coupling resistor. IDSEL mapping per P64H2 pin is arbitrary. However, AD16 is reserved.
Figure 8-6. IDSEL Sample Implementation Circuit
®
Intel
P64H2
IDSEL0
IDSEL1
IDSEL2
IDSEL3
100 Ω
100 Ω
100 Ω
100 Ω
PCI SLOT 1
PCI SLOT 2
PCI SLOT 3
PCI SLOT 4
8.1.6SMBus Address
The SMBus interface does not have configuration registers. The SMBus address is set by the states
of pins PA_GNT[5:4] and PB_GNT[5:4] when PWROK is asserted as described in Table 8-7.
Refer to the Intel
strap latching.
NOTE: There is no bit 0 because it is the read/write direction indicator.
®
PCI-64 Hub 2 (P64H2) Datasheet for a more detailed description of P64H2
98Design Guide
Intel® 82870P2 (P64H2)
8.2Hot Plug Implementation
The P64H2 contains two integrated Hot Plug Controllers (one per PCI/PCI-X interface) that
operate independently. These integrated controllers can be individually disabled or configured to
operate in one of the three defined modes of operation: Single Slot Parallel mode, Dual Slot
Parallel mode, and Serial mode. This section describes each of these three modes of operation, as
well as switch and button implementation and the Hot Plug Standard Usage Model.
8.2.1Stan dard Usage Model
To define a programming model for the Hot Plug Controllers (HPC), it is necessary to make some
assumptions about the interface between a user and a Hot Plug system that must be incorporated
into the hardware solution. The programming model includes two LED indicators, one optional
push button, and a sensor on the manually-operated retention latch (MRL) for each supported slot.
See Section 8.2.2 for MRL and attention button implementation. Section 8.2.3 describes the LED
indicators. For more information on the standard usage model, see the PCI Standard Hot -Plug
Controller and Subsystem Specification, Revision 1.0.
Caution: Users must always notify the operating system via a software user interf ace or Attention B utton (if
present) before opening an MRL. This allows the operating system to isolate the slot from the PCI
bus and unload the device driver gracefully. The unexpected opening of an MRL leads to
unpredictable results, including data corruption, abnormal termination of the operating system, or
damage to card or platform hardware.
8.2.1.1Hot-Removals
1. User selects a slot holding an enabled add-in card and requests that slot be disabled.
a. User interacts with a software user interface to request that slot be disabled.
b. User confirms request. System software validates request and initiates slot power down
sequence. Power Indicator LED blinks.
a. User presses momentary Attention Button at that slot.
b. Software interprets change on HxPRSNT# pin as a push button event. (Software ignores
second interrupt on HxPRSNT# caused by button release.) Power Indicator LED blinks.
c. User is permitted to cancel request within 5 seconds by pressin g Attention Button again.
d. System software validates request and initiates slot power down sequence.
2. System software waits for card activity on the PCI bus to end.
3. Hot Plug Controller asserts RST#, bus signals and clock lines are disconnected from the slot,
and power is removed.
4. Power Indicator LED is turned off. User may open MRL, disconnect cables, and remove card.
-- OR --
Design Guide99
Intel® 82870P2 (P64H2)
8.2.1.2Hot-Insertions
1. User selects an empty, disabled slot and opens MRL.
2. User inserts add-in card, closes MRL, and attaches cables to card.
3. User requests that slot be enabled.
a. User requests that slot be enabled via a software user interface.
b. Power Indicator LED next to slot blinks while system software validates request.
a. User presses momentary Attention Button at that slot.
b. Software interprets change on HxPRSNT# pin as a push button event. (Software ignores
second interrupt on HxPRSNT# caused by button release.)
c. User is permitted to cancel request within 5 seconds by pressing Attention Button again.
d. Power Indicator LED next to slot blinks while system software validates request.
4. Hot Plug Controller asserts RST# to the slot; main supply voltages are present at the slot.
5. Clock and bus signals are connected to the slot; RST# is deasserted.
-- OR --
6. Power Indicator LED is turned on. The slot is ready for operation.
8.2.2Hot Plug Switch Implementation
The mechanical design for the chassis should include a manually-operated retention latch (MRL)
that holds an add-in card in the slot. Each MRL should have an associated switch, optical device, or
other type of sensor to indicate whether a slot is “opened” or “closed .” (Note that the terms op ened
and closed do not necessarily indicate the electrical state of the switches used, but should be
thought of as a mechanical door that enables or disables cards to be installed or removed.) A slot
can be auto-powered down should someone attempt to remove a card without first notifying the
operating system. The mechanical design sho uld be such that it is impossib le for an ex pansion card
to be removed without the switch indicating that the slot is open. The mechanical design should
also prevent inadvertent switch “openings.”
An Attention Button is a momentary-contact push-button. This button serves to invoke the hot-p lug
service so that an adapter can be added or removed without the use of a software interface. Suppo rt
for the Attention Button is optional.
100Design Guide
Loading...
+ hidden pages
You need points to download manuals.
1 point = 1 manual.
You can buy points or you can get point for every manual you upload.