AMD, the AMD Arrow logo, AMD PowerNow!, AMD Virtualization, AMD-V, and combinations thereof are trademarks of Advanced Micro Devices, Inc.
HyperTransport is a licensed trademark of the HyperTransport Technology Consortium.
Microsoft, Windows, and Windows Server are registered trademarks of Microsoft Corporation.
PCI Express and PCIe are registered trademarks of PCI-SIG.
Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.
Disclaimer
The contents of this document are provided in connection with Advanced Micro Devices, Inc. ("AMD") products. AMD makes no representations or warranties with
respect to the accuracy or completeness of the contents of this publication and reserves the right to make changes to specifications and product descriptions at any time
without notice. AMD assumes no liability whatsoever, and disclaims any express or implied warranty, relating to this document including, but not limited to, the implied
warranty of merchantability, fitness for a particular purpose, or infringement of any intellectual property right. AMD shall not be liable for any damage, loss, expense,
or claim of loss of any kind or character (including without limitation direct, indirect, consequential, exemplary, punitive, special, incidental or reliance damages) arising
from use of or reliance on this document. No license, whether express, implied, arising by estoppel, or otherwise, to any intellectual property rights are granted by this
publication. Except for AMD product purchased pursuant to AMD's Standard Terms and Conditions of Sale, and then only as express ly set forth therein, AMD's products
are not designed, intended, authorized or warranted for use as components in systems intended for surgical implant into the body, or in other applications intended to
support or sustain life, or in any other application in which the failure of AMD's product could create a situation where personal injury, death, or severe property or
environmental damage may occur. AMD reserves the right to discontinue or make changes to its products at any time without notice.
1.1 Introducing the SR5650 ......................................................................................................................................................1-1
1.2 SR5650 Features .................................................................................................................................................................1-1
1.2.3A-Link Express II Interface..................................................................................................................................1-1
1.2.4Multiple Processor Support ..................................................................................................................................1-2
1.2.5Multiple Northbridge Support ..............................................................................................................................1-2
1.2.6Power Management Features ...............................................................................................................................1-2
1.2.8Test Capability Features .......................................................................................................................................1-2
1.4 Device ID ............................................................................................................................................................................1-3
1.6 Conventions and Notations .................................................................................................................................................1-4
1.6.5Acronyms and Abbreviations ...............................................................................................................................1-5
2.1.2HyperTransport™ Flow Control Buffers .............................................................................................................2-3
2.4.2Non-SB IOAPIC Support .....................................................................................................................................2-4
2.5 RAS Features ......................................................................................................................................................................2-7
2.5.2SERR_FATAL# and NON_FATAL_CORR# Pins .............................................................................................2-7
2.5.3NMI# and SYNCFLOODIN# ..............................................................................................................................2-8
2.5.4Suggested Platform Level RAS Sideband Signal Connections............................................................................2-8
2.5.5Error Reporting and Logging ...............................................................................................................................2-9
2.5.6Interrupt Generation on Errors ........................................................................................................................... 2-11
2.5.7Poisoned Data Support ....................................................................................................................................... 2-11
2.5.8PCIe® Link Disable State ..................................................................................................................................2-11
2.5.9HT Syncflood Based on PCIe® Error ............................................................................................................... 2-12
3.4.1PCI Express® Interface for General Purpose External Devices ......................................................................... 3-5
3.4.2A-Link Express II Interface to Southbridge ........................................................................................................ 3-5
3.6 Power Management Pins.................................................................................................................................................... 3-6
3.8 Power Pins.......................................................................................................................................................................... 3-7
4.5 Power Rail Sequence.......................................................................................................................................................... 4-2
4.5.1Power Up ............................................................................................................................................................. 4-3
4.5.2Power Down ........................................................................................................................................................ 4-4
Chapter 5: Electrical Characteristics and Physical Data
5.1.1Maximum and Minimum Ratings........................................................................................................................ 5-1
5.3 Package Information .......................................................................................................................................................... 5-4
5.3.2Board Solder Reflow Process Recommendations ............................................................................................... 5-6
Chapter 6: Power Management and ACPI
6.1 ACPI Power Management Implementation ....................................................................................................................... 6-1
Chapter 7: Testability
7.1 Test Capability Features..................................................................................................................................................... 7-1
7.2 Test Interface...................................................................................................................................................................... 7-1
7.3 XOR Tree ........................................................................................................................................................................... 7-1
7.3.1Brief Description of an XOR Tree .......................................................................................................................7-1
7.3.2Description of the XOR Tree for the SR5650 ......................................................................................................7-2
7.3.3XOR Tree Activation ...........................................................................................................................................7-2
7.3.4XOR Tree for the SR5650....................................................................................................................................7-3
7.4.1Brief Description of a VOH/VOL Tree................................................................................................................7-4
7.4.2VOH/VOL Tree Activation..................................................................................................................................7-5
7.4.3VOH/VOL pin list ................................................................................................................................................7-6
Appendix A: Pin Listings
7.5 SR5650 Pin Listing Sorted by Ball Reference................................................................................................................... A-2
A.1 SR5650 Pin Listing Sorted by Pin Name .......................................................................................................................... A-9
Figure 4-1: SR5650 Power Rail Power Up Sequence .................................................................................................................... 4-3
Figure 7-1: XOR Tree .................................................................................................................................................................... 7-2
Figure 7-2: Sample of a Generic VOH/VOL Tree ......................................................................................................................... 7-5
Table 1-1: Device IDs for the SR5690/5670/5650 Chipset Family ................................................................................................1-3
Table 1-2: Pin Type Codes ..............................................................................................................................................................1-4
Table 1-3: Acronyms and Abbreviations ........................................................................................................................................1-5
Table 2-1: SR5650 HyperTransport™ Flow Control Buffers .........................................................................................................2-3
Table 2-2: Types of Errors Detectable by the SR5650 AER Implementation ..............................................................................2-10
Table 2-3: Types of HyperTransport™ Errors Supported by the SR5650 ....................................................................................2-11
Table 2-4: Possible Configurations for the PCI Express® General Purpose Links ......................................................................2-12
Table 3-2: PCI Express® Interface for General Purpose External Devices ....................................................................................3-5
Table 3-3: 1 x 4 Lane A-Link Express II Interface for Southbridge ...............................................................................................3-5
Table 3-8: Power Pins .....................................................................................................................................................................3-7
Table 3-10: Strap Definitions for the SR5650 ..............................................................................................................................3-10
Table 3-11: Strap Definition for STRAP_PCIE_GPP_CFG .........................................................................................................3-10
Table 4-1: Timing Requirements for PCIe® Differential Clocks (GPP1_REFCLK and GPP3_REFCLK at 100MHz) ...............4-1
Table 4-2: Timing Requirements for HyperTransport™ Reference Clock (100MHz) ...................................................................4-1
Table 4-3: Timing Requirements for OSCIN Reference Clock (14.3181818MHz) .......................................................................4-2
Table 4-4: Power Rail Groupings for the SR5650 ..........................................................................................................................4-2
Table 4-5: SR5650 Power Rail Power-up Sequence .......................................................................................................................4-3
Table 5-1: Power Rail Maximum and Minimum Voltage Ratings .................................................................................................5-1
Table 5-2: Power Rail Current Ratings ...........................................................................................................................................5-1
Table 5-1: DC Characteristics for PCIe® Differential Clocks (GPP1_REFCLK and GPP3_REFCLK at 100MHz) ....................5-1
Table 5-3: DC Characteristics for 1.8V GPIO Pads ........................................................................................................................5-2
Table 5-4: DC Characteristics for the HyperTransport™ 100MHz Differential Clock (HT_REFCLK) .......................................5-2
Table 6-1: ACPI States Supported by the SR5650 .........................................................................................................................6-1
Table 7-1: Pins on the Test Interface ..............................................................................................................................................7-1
Table 7-2: Example of an XOR Tree ..............................................................................................................................................7-2
Table 7-3: SR5650 XOR Tree .........................................................................................................................................................7-3
Table 7-4: Truth Table for the VOH/VOL Tree Outputs ................................................................................................................7-5
The SR5650 is the system logic of the latest server/workstation platform from AMD that enables its next generation
CPUs. The SR5650 has a total of 26 PCI Express
are dedicated for the A-Link Express II interface to AMD’s Southbridges such as the SP5100 (formerly SB700S). The
SR5650 also comes equipped with the new HyperTransport™ 3 and PCIe Gen 2 technologies. All of these are achieved
by a highly integrated, thermally efficient design in a 29mm x 29mm package.
The SR5650 introduces a variety of Reliability, Availability and Serviceability (RAS) capabilities. These include parity
protection for on-chip memories, PCI Express Advanced Error Reporting (AER), and advanced error handling
capabilities for HyperTransport.
The SR5650 also supports a revision 1.26 compliant IOMMU (Input/Output Memory Management Unit) implementation
for address translation and protection services. This feature allows virtual addresses from PCI Express endpoint devices to
be translated to physical memory addresses. On-chip caching of address translations is provided to improve I/O
performance. The device is also compliant with revision 1.0 of the PCI Express Address Translation Services (ATS)
specification to enable ATS-compliant endpoint devices to cache address translation. These features enhance memory
protection and support hardware-based I/O virtualization when combined with appropriate operating system or hypervisor
software. Combined with AMD Virtualization™ (AMD-V™) technology, these features are designed to provide
comprehensive platform level virtualization support.
1.2SR5650 Features
1.2.1 CPU Interface
•
Supports 16-bit up/down HyperTransport™ (HT) 3.0 interface up to 5.2 GT/s.
•Supports 200, 400, 600, 800, and 1000 MHz HT1 frequencies.
•Supports 1200, 1400, 1600, 1800, 2000, 2200, 2400, and 2600 MHz HT3 frequencies (up to 2400 MHz only for the
RX980) .
•Supports “Shanghai” and subsequent series of AMD server/workstation and desktop processors through sockets F,
AM3, G34, and C32.
•Supports LDTSTOP interface and CPU throttling.
®
(PCIe®) lanes: 22 lanes are dedicated for external PCIe devices, and 4
1.2.2 PCI Express® Interface
•
Supports PCIe Gen 2 (version 2.0).
•Optimized peer-to-peer and general purpose link performance.
•Supports 22 PCIe Gen 2 general purpose lanes, and up to 8 devices on specific ports (possible configurations are
described in Section 2.6, “PCI Express®”).
•Supports a revision 1.26 compliant IOMMU (Input/Output Memory Management Unit) implementation for address
translation and protection services. Please refer to the AMD I/O Virtualization Technology (IOMMU) Specification
for more details.
1.2.3 A-Link Express II Interface
•
One x4 A-Link Express II interface for connection to an AMD Southbridge. The A-Link Express II is a proprietary
interface developed by AMD based on the PCI Express technology, with additional Northbridge-Southbridge
messaging functionalities.
Note: Branding can be in laser, ink, or
mixed laser-and-ink marking.
1.4Device ID
The SR5650 is a member of the AMD chipset family, which consists of different devices designed to support different
platforms. Each device is identified by a device ID, which is stored in the NB_DEVICE_ID register. The device IDs for
the SR5650/5690/5670 chipset family are as follows:
Table 1-1 Device IDs for the SR5690/5670/5650 Chipset Family
Note: Branding can be in laser, ink, or
mixed laser-and-ink marking.
Figure 1-3 SR5650 Alternate Branding for A21 Production ASIC (Lead Free Part)
1.6Conventions and Notations
Conventions and Notations
The following sections explain the conventions used throughout this manual.
1.6.1 Pin Names
Pins are identified by their pin names or ball references. All active-low signals are identified by the suffix ‘#’ in their
names (e.g., SYSRESET#).
1.6.2 Pin Types
The pins are assigned different codes according to their operational characteristics. These codes are listed in Table 1-2.
Table 1-2 Pin Type Codes
CodePin Type
IDigital Input
ODigital Output
I/OBi-Directional Digital Input or Output
MMultifunctional
PwrPower
GndGround
A-OAnalog Output
A-IAnalog Input
A-I/OAnalog Bi-Directional Input/Output
A-PwrAnalog Power
A-GndAnalog Ground
OtherPin types not included in any of the categories above
1.6.3 Numeric Representation
Hexadecimal numbers are appended with “h” whenever there is a risk of ambiguity. Other numbers are in decimal.
Pins of identical functions but different trailing digits (e.g., DFT_GPIO0, DFT_GPIO1, ...DFT_GPIO5) are referred to
collectively by specifying their digits in square brackets and with colons (i.e., “DFT_GPIO[5:0]”). A similar short-hand
notation is used to indicate bit occupation in a register. For example, NB_COMMAND[15:10] refers to the bit positions
10 through 15 of the NB_COMMAND register.
Phrases or sentences in blue italicfont are hyperlinks to other parts of the manual. Users of the PDF version of this manual
can click on the links to go directly to the referenced sections, tables, or figures.
1.6.5 Acronyms and Abbreviations
The following is a list of the acronyms and abbreviations used in this manual.
Table 1-3 Acronyms and Abbreviations
AcronymFull Expression
ACPIAdvanced Configuration and Power Interface
ASPMActive State Power Management
A-Link-EA-Link Express interface between the Northbridge and Southbridge.
BGABall Grid Array
BIOS
BISTBuilt In Self Test.
DBIDynamic Bus Inversion
DPMDefects per Million
EPROMErasable Programmable Read Only Memory
FCBGAFlip Chip Ball Grid Array
FIFOFirst In, First Out
VSSGround
GPIOGeneral Purpose Input/Output
HTHyperTransport™ interface
IDDQDirect Drain Quiescent Current
IOMMUInput/Output Memory Management Unit
JTAGJoint Test Access Group. An IEEE standard.
MBMega Byte
NBNorthbridge
PCIPeripheral Component Interface
®
PCIe
PLLPhase Locked Loop
POSTPower On Self Test
PDPull-down Resistor
PUPull-up Resistor
RASReliability, Availability and Serviceability
SBSouthbridge
TBATo Be Added
VRMVoltage Regulation Module
Basic Input Output System. Initialization code stored in a ROM or Flash RAM used to start up a
system or expansion card.
This chapter describes the functional operation of the major interfaces of the SR5650 system logic chip. Figure 2-1
illustrates the SR5650 internal blocks and interfaces.
Figure 2-1 SR5650 Internal Blocks and Interfaces
2.1HyperTransport™ Interface
2.1.1 Overview
The SR5650 is optimized to interface with “Shanghai” and subsequent series of AMD server/workstation and desktop
processors through sockets F, AM3, G34, and C32. The SR5650 supports HyperTransport
HyperTransport 1 (HT1) for backward compatibility and for initial boot-up. For a detailed description of the interface,
please refer to the HyperTransport I/O Link Specification from the HyperTransport Consortium. Figure 2-2,
“HyperTransport™ Interface Block Diagram,” illustrates the basic blocks of the host bus interface of the SR5650.
The SR5650 HyperTransport bus interface consists of 16 unidirectional differential Command/Address/Data pins, and 2
differential Control pins and 2 differential Clock pins in both the upstream and downstream directions. On power up, the
link is 8-bit wide and runs at a default speed of 400MT/s in HyperTransport 1 mode. After negotiation, carried out by the
HW and SW together, the link width can be brought up to the full 16-bit width and the interface can run up to 5.2GT/s in
HyperTransport 3 mode. In HyperTransport 1 mode, the interface operates by clock-forwarding while in HyperTransport
3 mode, the interface operates by dynamic phase recovery, with frequency information propagated over the clock pins.
The interface is illustrated below in Figure 2-3, “SR5650 HyperTransport™ Interface Signals.” The signal name and
direction for each signal is shown with respect to the SR5650. Detailed descriptions of the signals are given in Section 3.3,
The SR5650’s IOMMU (Input/Output Memory Management Unit) block provides address translation and protection
services as described in version 1.26 of the AMD I/O Virtualization Technology (IOMMU) Specification. The SR5650
also supports the PCI Express Address Translation Services 1.0 Specification, which allows the supporting of endpoint
devices to request and cache address translations.
When DMA requests containing virtual addresses are received, the IOMMU looks up the page translation tables located
in the system memory in order to convert the virtual addresses into physical addresses and to verify access privileges.
On-chip caching is provided in order to speed up translation and reduce or eliminate the number of system memory
accesses required. Every PCIe core contains a local translation cache, and the SR5650 also contains a shared global
translation cache.
The SR5650 supports up to 2
52-bit physical address space.
16
domains, each of which can utilize a separate 64-bit virtual address space. It supports a
2.3Multiple Northbridge Support
Multiple SR5690/5670/5650 (referred to as “SR56x0”below) Northbridges may be implemented in the same system given
enough free HyperTransport links from the processor complex. However, only a single Southbridge may be used. The
SR56x0 attached to the Southbridge is called the primary SR56x0, and any other instance of SR56x0 is called a secondary
SR56x0. The A-Link Express interface on any secondary SR56x0 must be left unconnected, and it cannot be used to
support any PCI Express endpoint devices.
IOMMU
The PWM_GPIO5 pin-strap is used to indicate whether an SR56x0 is a primary or a secondary Northbridge. If no
pull-down resistor is attached on the pin, the internal pull-up resistor on it will set the strap value to “1,” indicating the
device to be a primary Northbridge. On any secondary SR56x0, the PWM_GPIO5 pin-strap must be pulled low.
In the multi-NB mode, special PCI Express messages for functions such as PME may be passed from a secondary SR56x0
to the primary SR56x0 or the Southbridge over the HyperTransport bus. If the SR56x0’s internal IOAPIC is not used,
INTx messages may also be forwarded over the HyperTransport bus to the Southbridge IOAPIC. Peer-to-peer writes
between PCI Express endpoints are also allowed between any SR56x0 and another by routing peer-to-peer requests over
the HyperTransport bus.
Note: As it is possible to mix-and-match SR5650, SR5670, and SR5690 on the same system, whenever a
multiple-SR5650 configuration is being referred to in this document, it actually represents any combination of SR5650,
SR5670, and SR5690 possible under that situation. Some constrains may apply.
2.4Interrupt Handling
2.4.1 Legacy INTx Handling
In legacy interrupt mode, all INTx messages must be routed to the Southbridge IOAPIC. The primary NB directs all INTx
messages directly down to the Southbridge IOAPIC. Secondary NBs direct INTx messages up to the processor complex,
where they are broadcast down to all HT devices. See Section 2.3, “Multiple Northbridge Support‚’ on page 2-4 for
details.
The 4 legacy interrupts sent by endpoint devices (INT A/B/C/D) may undergo a 2-stage programmable swizzling process
that maps them onto the 8 possible internal INTx messages (INT A/B/C/D/E/F/G/H). The first swizzling stage is
performed by rotating the interrupt message number based upon the bridge device number. The second stage is register
controllable on a per-bridge basis and maps the rotated INT A/B/C/D onto INT E/F/G/H. INT A to H messages sent to the
Southbridge are mapped onto the SB IOAPIC interrupt redirection table entries 16 to 23.
2.4.2 Non-SB IOAPIC Support
The SR5650 supports routing legacy IOAPIC memory-mapped I/O addresses (0xFECx_xxxx) to any PCI Express port to
support endpoint devices with integrated IOAPIC.
The SR5650 supports routing local INTx messages to its integrated IOAPIC. The integrated IOAPIC contains a 32-entry
redirection table. INTx messages from endpoint devices, bridges, HTIU, and IOMMU can be mapped onto different
redirection entries under register control.
2.4.4 MSI Interrupt Handling and MSI to HT Interrupt Conversion
In MSI interrupt mode, all interrupts are sent directly from the endpoint devices through the SR5650 up to the processor
complex. All MSI interrupts are converted into HT-formatted interrupts. For MSIs from PCI Express endpoint devices
and internally generated PCI Express interrupts, the conversion occurs in the associated IOMMU L1 block. For IOMMU
interrupts and, optionally, HT error interrupts and internal parity error interrupts, the conversion occurs in the HTIU
block. HT error interrupts and internal parity error interrupts may be optionally redirected to an MSI generation block
underneath the SB VC1 IOMMU L1 so that they can be remapped by IOMMU. IOMMU internal MSI interrupts are never
remapped.
The PCI configuration spaces of each on-board device contains a fixed HT MSI mapping capability (except for Device 1,
which is unused). This implies that all MSI interrupts with address 0xFEEx_xxxx have to be converted to HT interrupts.
Because of this, software is required to program all MSI address registers with an 0xFEEx_xxxx address.
2.4.5 Internally Generated Interrupts
The SR5650 may internally generate interrupts for the following purposes:
•PCI Express error
•PCI Express PME
•HT error
•Internal parity error
•IOMMU command handler
•IOMMU event logger
Internally generated interrupts may be in either legacy INTx or MSI format. Internal MSI interrupt sources do not support
per-vector masking.
2.4.6 IOMMU Interrupt Remapping
When the IOMMU is enabled, interrupts generated downstream of the IOMMU are remapped based upon the IOMMU
tables. The following classes of interrupts are not remapped by the IOMMU because they are generated upstream of the
IOMMU:
•HT error (optional)
•Internal parity error (optional)
•IOMMU command handler and event logger
2.4.7 Interrupt Routing Architecture
2.4.7.1 Legacy Mode
Primary SR5650: Legacy INTx messages are routed directly to the SB IOAPIC. The SB IOAPIC generates upstream
interrupt requests, which are translated by the IOMMU before they are delivered up to the processor complex.
Secondary SR5650: Legacy INTx messages are routed over HyperTransport through the processor complex to the
primary SR5650, which forwards them to the SB IOAPIC. The SB IOAPIC generates upstream interrupt requests, which
are translated by the IOMMU before being delivered up to the processor complex.
The routing paths are illustrated in Figure 2-4 below.
INTx Message from device attached to primary SR5650
INTx Message from device attached to secondary SR5650
Interrupts from SB IOAPIC
CPU
SB
PCI-E
Endpoint
Device
IOMMU
HT
PCI-
Express
IOAPIC
SR5650
INTx Message from PCI-Express device attached to SR5650
Internal interrupt
Remapped HT Interrupt
SR5650
PCI-E
Endpoint
Device
CPU
Figure 2-4 Interrupt Routing Paths in Legacy Mode
2.4.7.2 Legacy Mode with Integrated IOAPIC
For both the primary and secondary SR5650s, legacy INTx messages are routed to the integrated IOAPICs of the
SR5650s, which generates interrupt requests. These requests are remapped by the IOMMU before being delivered up to
the processor complex. If an INTx message gets directed to an IOAPIC table entry that is not enabled, the IOAPIC sends
the INTx message back to the IOC to go to the SB PIC/IOAPIC.
The routing paths are illustrated in Figure 2-5 below.
Figure 2-5 Interrupt Routing Paths in Legacy Mode with Integrated IOAPIC
MSI Interrupt from PCI-Express device attached to SR5650
Remapped HT Interrupt
SR5650
PCI-E
Endpoint
Device
CPU
2.4.7.3 MSI Mode
For both the primary and secondary SR5650s: MSI interrupt requests are remapped by the IOMMU and sent up to the
processor complex. The routing path is illustrated in Figure 2-6 below.
2.5RAS Features
2.5.1 Parity Protection
All memories in SR5650 are parity protected to reduce the possibility of silent data corruption. Multiple parity words are
interleaved to convert burst errors (multiple physically adjacent bits corrupted) into multiple single-bit detectable errors to
increase robustness. The minimum number of interleaved parity words in any on-board memory is 4. All macros contain
2.5.1.1 Parity Protection for IOMMU Cache Memories
test circuitry for software to generate false errors on either the read or write side of the memory for verification of error
handling routines. Error injection circuitry only corrupts parity bits rather than real data bits to avoid data corruption.
All IOMMU cache memories are parity protected. When a parity error is detected, the access from the associated bank is
marked as an automatic miss. The cache line is marked as invalid and may later be overwritten with data from system
memory (which is ECC protected). The error is logged in a status bit and an optional interrupt is generated (either fatal,
non-fatal, or correctable parity error).
All normal memories are also parity protected. When a parity error is detected, the failure is likely to be fatal as there is no
automatic recovery mechanism and no way for hardware to tag a specific request or operation with the error. The error is
logged in a status bit for later diagnosis and an optional interrupt is generated (either fatal or non-fatal parity error).
The SR5650 implements a dedicated pin, DBG_GPIO0/SERR_FATAL#, to signal either a system or a fatal error, which
can be used to signal a BMC for further actions. SERR_FATAL# may be asserted on various error conditions like HT
Figure 2-6 Interrupt Routing Path in MSI Mode
syncflood, as well as internal parity errors or fatal errors for which signalling by SERR_FATAL# is enabled. Fatal errors
are identified via the fatal error status bits.
Non-fatal or correctable errors may be likewise signalled via DBG_GPIO3/NON_FATAL_CORR#.
The SERR_FATAL# and NON_FATAL_CORR# pin functionalities are disabled on warm reset.
2.5.3 NMI# and SYNCFLOODIN#
The SR5650 may configure the DFT_GPIO0/NMI# pin as an input pin for triggering an upstream NMI packet to the
processor complex. The pin should be driven by a BMC. An internal sticky status bit records the use of the NMI# pin.
Also, the SR5650 may configure the DFT_GPIO5/SYNCFLOODIN# pin as an input pin for triggering a HyperTransport
syncflood event. The pin should driven by a BMC. An internal sticky status bit records the use of the SYNCFLOODIN#
pin.
2.5.4 Suggested Platform Level RAS Sideband Signal Connections
Figure 2-7 is a logical diagram showing suggestions for RAS sideband signal connections at the platform level .
segments).
Alternately, this can
connect directly to a
BMC
SCL/SDA
Interrupt l ine to
Sys_SMBUS_IO_EXP_INTR_L
OPMA pin
Separate connections to
debug pins and GPIO
expander for each SR5650
in the system
NMI# only needs to
be connected on
the primary SR5650
Figure 2-7 Suggested Platform Level RAS Sideband Signal Connections
2.5.5 Error Reporting and Logging
2.5.5.1 PCI Error Logging
The SR5650 implements all PCI standard error logging bits for all on-board devices and functions including the host
bridge device, IOMMU, and PCI Express bridges.
2.5.5.2 PCIe
The SR5650 PCIe
®
Advanced Error Reporting
®
cores implement the optional Advanced Error Reporting (AER) feature mechanism in the PCI Express 2.0 Base Specification. Errors are logged for received packet errors such as poisoned data, malformed TLP, and
etc. within the PCIe core and are accessible via the bridge configuration spaces.
The ACS violations for ACS Source Validation and ACS Translation Blocking are recorded in the AER error log. Errors
due to IOMMU translation failures are not logged as ACS violations, but are logged as UR or CA depending on the error
type.
IOC may abort a non-posted request with UR status if it determines that the request will not hit system memory. Such
errors are pushed back into the PCIe core for logging. The IOC must abort potential peer-to-peer non-posted requests to
avoid a deadlock condition. For posted requests, the IOC can be configured to forward all non-decoded (non system
memory and non-peer-to-peer) posted requests up to the processor, which may abort the request and generate an MCA
error log.
For downstream completions with abort status coming back from the processor, error status is propagated to the endpoint
but no AER header information is logged in the chipset.
For upstream completions, error status is propagated up to the processor and AER information may be logged.
Table 2-2 lists the types of errors that are detectable by the SR5650 AER implementation. For details, see the PCI Express
2.0 Base Specification.
Table 2-2 Types of Errors Detectable by the SR5650 AER Implementation
Error TypeError Class
ACS ViolationUncorrectable – Fatal or Non-fatal
Unsupported RequestUncorrectable – Fatal or Non-fatal
Malformed TLPUncorrectable – Fatal or Non-fatal
Unexpected CompletionUncorrectable – Fatal or Non-fatal
Completer AbortUncorrectable – Fatal or Non-fatal
Completion TimeoutUncorrectable – Fatal or Non-fatal
Poisoned TLP ReceivedUncorrectable – Fatal or Non-fatal
Data Link Layer Protocol ErrorUncorrectable – Fatal or Non-fatal
ECRC ErrorUncorrectable – Fatal or Non-fatal
Replay TimeoutCorrectable
REPLAY_NUM RolloverCorrectable
Bad DLLPCorrectable
Bad TLPCorrectable
The following error classes are NOT supported:
•Receiver Overflow Error
•Flow Control Error
•Surprise Down Error
•Receiver Error
2.5.5.3 IOMMU Error Reporting
The IOMMU specification defines a standard error logging facility that logs error events in system memory with register
status bits or interrupt notification to system software. The SR5650 fully supports the generation of logging events
following this standard.
2.5.5.4 HyperTransport™ Error Reporting
The HyperTransport specification defines various levels of error handling for link-related errors. The SR5650 supports
the detection of most error classes including protocol error, overflow error, and response error. The SR5650 also supports
notification of error conditions via fatal interrupts, non-fatal interrupts, or syncflood.
Table 2-3 lists the types of errors supported by the error handling capabilities of the SR5650 for HyperTransport.
Protocol ErrorProtocol conditions detected in HT1 mode:
received data did not match size of requested data.
mode. In HT3 mode, this maps onto a retry in the hope that when the packet is subsequently received,
there is space in the FCB. No interrupt will be generated in HT3 mode.
• Data count not matching header
• Invalid command encoding
• Invalid CTL encoding
• Incomplete header
• Unexpected data
Protocol conditions detected in HT3 mode
• Data count not matching header
• Invalid command encoding
• Invalid CTL encoding
• Incomplete header
• Unexpected data
• Unexpected CRC
• Missing CRC
• Non-NOP inserted command
• Inserted command without inserted command CTL encoding
End of chain error is not supported, since the end of the chain is on PCI Express instead HyperTransport.
2.5.5.5 Internal Parity Error Reporting
One register bit per memory macro is used to log parity errors. Values for those bits are persistent across a warm reset for
diagnostic purposes.
2.5.6 Interrupt Generation on Errors
Internal interrupts may be generated on the following error conditions:
•PCI Express errors (fatal, non-fatal, or correctable)
•HT errors (fatal or non-fatal)
•IOMMU events
•Internal parity error (fatal or non-fatal)
•Internal parity error in the IOMMU cache (fatal, non-fatal, or correctable)
2.5.7 Poisoned Data Support
The SR5650 supports the propagation of poisoned data attributes (EP in PCIe and Data Error in HT) between PCI Express
endpoints and the processor for both host and DMA requests or responses. The SR5650 cannot actively mark a transaction
with a poisoned data attribute even if the transaction encounters an internal parity error. Received packets containing
ECRC errors are not marked as poisoned.
2.5.8 PCIe® Link Disable State
The SR5650 has the ability to put PCIe links into the disabled state as an error response in order to help stop data
movement within the system. Links which received fatal errors may be disabled. Also, a HyperTransport syncflood event
may be used to trigger all links to enter the disabled state.
The SR5650 has the ability to put the HyperTransport link into the syncflood state when a fatal or non-fatal error is
received on the PCIe interface. This is done in order to help stop data movement within the system.
PCI
Express®
2.6PCI Express
2.6.1 PCIe® Ports
In total, there are 9 PCIe® ports on the SR5650, divided into 3 groups and implemented in hardware as 3 separate cores:
•PCIE-GPP1: 2 general purpose ports, 16 lanes in total. Width of each port is x8. In the default configuration, the 2
ports are combined to provide a 1 x16 port.
•PCIE-GPP3: 6 general purpose ports, with 6 lanes in total. They support 6 different configurations with respect to
link widths: 4:2, 4:1:1, 2:2:2, 2:2:1:1, 2:1:1:1:1, and 1:1:1:1:1:1 (default configuration). For details on the possible
configurations for the GPP3 lanes, see Table 2-4 below and .
Table 2-4 Possible Configurations for the PCI Express® General Purpose Links
PCIe CorePhysical LaneConfig. BConfig. C Config. C2 Config. EConfig. KConfig. L
GPP3
•PCIE-SB: The Southbridge port provides a dedicated x4 link to the Southbridge (also referred to as the “A-Link
Express II interface”).
Each port supports the following PCIe functions:
®
GPP3 lane 0
GPP3 lane 1x1
GPP3 lane 2
GPP3 lane 3x1x1
GPP3 lane 4
GPP3 lane 5x1x1x1x1
x4x4
x2
x1
x2x2x2
x2
x2
x1
x1x1x1
x2
x1
x1
•PCIe Gen 1 link speeds
•ASPM L0s and L1 states
•ACPI power management
•Endpoint and root complex initiated dynamic link degradation
•Lane reversal
•Alternative Routing-ID Interpretation (ARI)
•Access Control Services (ACS)
•Advanced Error Reporting (AER)
•Address Translation Services (ATS)
2.6.2 PCIe® Reset Signals
Reset signals to PCIe slots, as well as embedded PCIe devices, must be controlled through one or more
software-controllable GPIO pins instead of the global system reset. It is recommended that unique GPIO pins be used for
each slot or device. The SR5650 has four GPIO pins that may be used for the purpose of driving reset signals
(PCIE_GPIO_RESET[5:4] and PCIE_GPIO_RESET[2:1]). Additional reset GPIO pins may be driven by
platform-specific means such as a super I/O or an I/O expander.
2.7External Clock Chip
On the SR5650 platform, an external clock chip provides the CPU, PCI Express, and A-Link Express II reference clocks.
For requirements on the clock chip, please refer to the 800-Series IGP Express AMD Platform External Clock Generator Requirements Specification for Server Platforms.
This chapter gives the pin descriptions and the strap options for the SR5650. To jump to a topic of interest, use the
following list of hyperlinked cross references:
“Pin Assignment Top View” on page 3-2
“SR5650 Interface Block Diagram” on page 3-4
“CPU HyperTransport™ Interface” on page 3-4
“PCI Express® Interfaces” on page 3-5:
“PCI Express® Interface for General Purpose External Devices” on page 3-5
“A-Link Express II Interface to Southbridge” on page 3-5
Figure 3-1 shows the different interfaces on the SR5650. Interface names in blue are hyperlinks to the corresponding
sections in this chapter.
SR5650 Interface Block Diagram
Figure 3-1 SR5650 Interface Block Diagram
3.3CPU HyperTransport™ Interface
Table 3-1 HyperTransport™ Interface
Pin NameType
HT_RXCAD[15:0]P,
HT_RXCAD[15:0]N
HT_RXCLK[1:0]P,
HT_RXCLK[1:0]N
HT_RXCTL[1:0]P,
HT_RXCTL[1:0]N
HT_TXCAD[15:0]P,
HT_TXCAD[15:0]N
IVDDHTVSSReceiver Command, Address, and Data Differential Pairs
IVDDHTVSS
IVDDHTVSS
OVDDHTVSSTransmitter Command, Address, and Data Differential Pairs
Power
Domain
Ground
Domain
Functional Description
Receiver Clock Signal Differential Pair. Forwarded clock signal. Each byte of
RXCAD uses a separate clock signal. Data is transferred on each clock edge.
Receiver Control Differential Pair. The pair is for distinguishing control packets
from data packets. Each byte of RXCAD uses a separate control signal.
HT_RXCALNOtherVDDHTVSSReceiver Calibration Resistor to HT_RXCALP
HT_RXCALPOtherVDDHTVSSReceiver Calibration Resistor to HT_RXCALN
HT_TXCALPOtherVDDHTVSSTransmitter Calibration Resistor to HTTX_CALN
HT_TXCALNOtherVDDHTVSSTransmitter Calibration Resistor to HTTX_CALP
OVDDHTVSS
OVDDHTVSS
Power
Domain
Ground
Domain
Functional Description
Transmitter Clock Signal Differential Pair. Forwarded clock signal. Each byte
of TXCAD uses a separate clock signal. Data is transferred on each clock
edge.
Transmitter Control Differential Pair. The pair is for distinguishing control
packets from data packets. Each byte of TXCAD uses a separate control
signal.
3.4PCI Express® Interfaces
3.4.1PCI Express® Interface for General Purpose External Devices
Table 3-2 PCI Express® Interface for General Purpose External Devices
Pin NameType
GPP1_TX[15:0]P,
GPP1_TX[15:0]N
GPP1_RX[15:0]P,
GPP1_RX[15:0]N
GPP3_TX[5:0]P,
GPP3_TX[5:0]N
GPP3_RX[5:0]P,
GPP3_RX[5:0]N
OVDDA18PCIE VSSA_PCIE
OVDDA18PCIE VSSA_PCIE
Power
Domain
IVDDA18PCIE VSSA_PCIE
IVDDA18PCIE VSSA_PCIE
Ground
Domain
Integrated
Termination
50 between
complements
50 between
complements
50 between
complements
50 between
complements
Functional Description
General Purpose 1 Transmit Data Differential Pairs.
Connect to connector[s] for general purpose external
device[s] on the motherboard.
General Purpose 1 Receive Data Differential Pairs.
Connect to connector[s] for general purpose external
device[s] on the motherboard.
General Purpose 3 Transmit Data Differential Pairs.
Connect to connector[s] for general purpose external
device[s] on the motherboard.
General Purpose 3 Receive Data Differential Pairs.
Connect to connector[s] for general purpose external
device[s] on the motherboard.
3.4.2 A-Link Express II Interface to Southbridge
Table 3-3 1 x 4 Lane A-Link Express II Interface for Southbridge
Southbridge Transmit Data Differential Pairs. Connect to the
corresponding Receive Data Differential Pairs on the
Southbridge.
Southbridge Receive Data Differential Pairs. Connect to the
corresponding Transmit Data Differential Pairs on the
Southbridge.
3.4.3 Miscellaneous PCI Express® Signals
Table 3-4 Miscellaneous PCI Express® Signals
Clock Interface
Pin NameType
PCE_BCALRNIVDDA18PCIEVSSA_PCIE
PCE_BCALRPIVDDA18PCIEVSSA_PCIE
PCE_TCALRNIVDDA18PCIEVSSA_PCIE
PCE_TCALRPIVDDA18PCIEVSSA_PCIE
PCE_RCALRNIVDDA18PCIEVSSA_PCIE
PCE_RCALRPIVDDA18PCIEVSSA_PCIE
PCIE_RESET_GP
IO[5:1]
3.5Clock Interface
Table 3-5 Clock Interface
Pin NameType
HT_REFCLKP,
HT_REFCLKN
GPP1_REFCLKP,
GPP1_REFCLKN
GPP3_REFCLKP,
GPP3_REFCLKN
OSCINIVDD18VSSDisabled
Power
Domain
I/OVDDA18PCIEVSS
Power
Domain
IVDDA18HTPLLVSSA_HTDisabled
IVDDA18PCIEVSSA_PCIE–
IVDDA18PCIEVSSA_PCIE–
Ground
Domain
Ground
Domain
Functional Description
N Channel Driver Compensation Calibration for Rx and Tx Channels
on Bottom Side.
P Channel Driver Compensation Calibration for Rx and Tx Channels
on Bottom Side
N Channel Driver Compensation Calibration for Rx and Tx Channels
on Top Side.
P Channel Driver Compensation Calibration for Rx and Tx Channels
on Top Side
N Channel Driver Compensation Calibration for Rx and Tx Channels
on Right Side.
P Channel Driver Compensation Calibration for Rx and Tx Channels
on Right Side
PCIe Resets. Except for PCIE_RESET_GPIO3, they can also be
used as GPIOs. There are internal pull-downs of 1.7 k on these
pins.
Integrated
Termination
Functional Description
HyperTransport™ 100 MHz Clock Differential Pair from
external clock source
General Purpose 1 Clock Differential Pair. The pair is
connected to an external clock generator on the
motherboard when the General Purpose 1 link is used,
and can be left unconnected if the link is not used.
General Purpose 3 Clock Differential Pair. The pair has to
be connected to an external clock generator on the
motherboard whether the General Purpose 3 link is
used or not.
14.318MHz Reference clock input from the external clock
chip (1.8 volt signaling)
3.6Power Management Pins
Table 3-6 Power Management Pins
Pin NameType
ALLOW_LDTSTOPODVDD18VSSAllow LDTSTOP. This signal is used by the SR5650 to communicate with the
LDTSTOP#IVDD18VSSHyperTransport™ Stop. This signal is generated by the Southbridge and is used to
POWERGOOD IVDD18VSS
SYSRESET#IVDD18VSSGlobal Hardware Reset. This signal comes from the Southbridge.
Southbridge and tell it when it can assert the LDTSTOP# signal.
1 = LDTSTOP# can be asserted
0 = LDTSTOP# has to be de-asserted
determine when the HyperTransport link should be disconnected and go into a
low-power state. It is a single-ended signal.
Input from the motherboard signifying that the power to the SR5650 is up and
ready. Signal High means all power planes are valid. It is not observed internally
until it has been high for more than 6 consecutive REFCLK cycles. The rising edge
of this signal is deglitched.
Miscellaneous Pins
3.7Miscellaneous Pins
Table 3-7 Miscellaneous Pins
Pin NameType
I2C_CLKI/OVDD18VSS–I2C interface clock signal. Can also be used as GPIO.
I2C_DATAI/OVDD18VSS–I
STRP_DATAI/OVDD18VSS–
TESTMODEIVDD18VSS–
DFT_GPIO5/
SYNCFLOODIN#
DFT_GPIO[4:1],I/OVDD18VSSPull Up
DFT_GPIO0/NMI#I/OVDD18VSSPull Up
DBG_GPIO3/
NON_FATAL_CORR#
DBG_GPIO0/
SERR_FATAL#
THERMALDIODE_P,
THERMALDIODE_N
A-O–––
Power
Domain
I/OVDD18VSSPull Up
I/OVDD18VSSPull Up
I/OVDD18VSSPull Up
Ground
Domain
Integrated
Termination
Functional Description
2
C interface data signal. Can also be used as GPIO.
2
I
C interface data signal for external EEPROM based strap loading.
See the SR5650 Strap Document for details on the operation.
When High, puts the SR5650 in test mode and disables the
SR5650 from operating normally.
Output for DFT TESTMODE, or Syncflood input for triggering a
HyperTransport™ syncflood event.
Because the pin is used as a pin strap during the power-on of the
SR5650, an external device must not drive the pin until after
SYSRESET# is deasserted. Also, the pin is not 3.3V tolerant and
needs a level shifter when interfacing to a 3.3V line.
The pin cannot be used for general GPIO functions.
Outputs for DFT TESTMODE. These pins cannot be used for
general GPIO functions.
Output for DFT TESTMODE, or NMI input for triggering an
upstream NMI packet to the processor complex.
Because the pin is used as a pin strap during the power-on of the
SR5650, an external device must not drive the pin until after
SYSRESET# is deasserted. Also, the pin is not 3.3V tolerant and
needs a level shifter when interfacing to a 3.3V line.
The pin cannot be used for general GPIO functions.
Output for Debug Bus, or Non-Fatal or Correctable Error signal to
BMC. The pin is not 3.3V tolerant and needs a level shifter when
interfacing to a 3.3V line. When used as a debug bus output, the
pin’s NON_FATAL_CORR# function is overridden.
The pin cannot be used for general GPIO functions.
Output for Debug Bus, or System Error or Fatal Error signal to
BMC. The pin is not 3.3V tolerant and needs a level shifter when
interfacing to a 3.3V line. When used as a debug bus output, the
pin’s SERR_FATAL# function is overridden.
The pin cannot be used for general GPIO functions.
Diode connections to external SM Bus microcontroller for
monitoring IC thermal characteristics.
3.8Power Pins
Table 3-8 Power Pins
Pin NameVoltage
VDDC1.1V18
VDD181.8V5A18, B18, C18, D18, E18I/O Power for GPIO pads
The SR5650 provides strapping options to define specific operating parameters. The strap values are latched into internal
registers after the assertion of the POWERGOOD signal to the SR5650. Table 3-10, “Strap Definitions for the SR5650,”
shows the definitions of all the strap functions. These straps are set by one of the following four methods:
•Allowing the internal pull-up resistors to set all strap values “1”’s automatically.
•Attaching pull-down resistors to specific strap pins listed in Table 3-10 to set their values to “0”’s.
•Downloading the strap values from an I
representative for details).
•Setting through an external debug port, if implemented (contact your AMD FAE representative for details).
Table 3-10 Strap Definitions for the SR5650
Strap FunctionStrap PinDescription
PRIMARY_NBPWM_GPIO5Indicates whether the device is a primary or a secondary Northbridge on a
ReservedPWM_GPIO[4:2]Reserved. Make provision for an external pull-down resistor on each of the pins, but do
ReservedDFT_GPIO0/NMI# Reserved. Make provision for an external pull-down resistor on this pin, but do not install
LOAD_ROM_STRAPS#DFT_GPIO1Selects loading of strap values from EEPROM
STRAP_PCIE_GPP_CFGDFT_GPIO[4:2]General Purpose Link 3 Configuration.
ReservedDFT_GPIO5/
SYNCFLOODIN#
Strapping Options
2
C serial EEPROM (for debug purpose only; contact your AMD FAE
multiple-Northbridge platform. See
on page 2- 4
0: Device is a secondary Northbridge
1: Device is the primary Northbridge (Default)
not install a resistor.
a resistor.
2
0: I
C master can load strap values from EEPROM if connected, or use hardware
default values if not connected
1: Use hardware default values (Default)
See
Table 3-11 below for details.
Reserved. Make provision for an external pull-down resistor on this pin, but do not install
a resistor.
for details. Do not install a resistor for sinlge-Northbridge platforms.
section 2.3, “Multiple Northbridge Support,”
Table 3-11 Strap Definition for STRAP_PCIE_GPP_CFG
Strap Pin ValueLink Width
GPP3
GPP3
GPP3
GPP3
GPP3
DFT_GPIO4 DFT_GPIO3 DFT_GPIO2
111
110
101x2x2x2C2
100x2x2x1x1K
011x2x1x1x1x1E
010x1x1x1x1x1x1
001x4x1x1C
000x4x2B
Note: If the pin straps instead of strap values from EEPROM are used, the GPP3
configuration will then be determined according to this table and cannot be changed after
the system has been powered up.
Table 4-2 Timing Requirements for HyperTransport™ Reference Clock (100MHz) (Continued)
SymbolParameterMinimumMaximumUnitNote
DCDuty Cycle4555%11
Notes:
More details are available in AMD HyperTransport 3.0 Reference Clock Specification and AMD Family 10h Processor Reference Clock
Parameters, document # 34864
1 Single-ended measurement at crossing point. Value is maximum-minimum over all time. DC Value of common mode is not important
due to blocking cap.
2 Minimum frequency is a consequence of 0.5% down spread spectrum.
3 Measured with spread spectrum turned off.
4 Only simulated at the receive die pad. This parameter is intended to give guidance for simulation. It cannot be tested on a tester but is
guaranteed by design.
5 Differential measurement through the range of ±100mV, differential signal must remain monotonic and within slew rate specification
when crossing through this region.
6 T
7 Accumulated T
8 V
9 V
V
10 The difference in magnitude of two adjacent V
11 Defined as t
is the maximum difference of t
jc max
D(PK-PK)
is the amplitude of the ring-back differential measurement, guaranteed by design that the ring-back will not cross 0V VD.
D(min)
is the largest amplitude allowed.
D(max)
over a 10s time period, measured with JIT2 TIE at 50ps interval.
jc
is the overall magnitude of the differential signal.
HIGH/tCYCLE
between any two adjacent cycles.
CYCLE
measurements. V
DDC
4.4OSCIN Reference Clock Timing Parameters
OSCIN Reference Clock Timing Parameters
is the stable post overshoot and ring-back part of the signal.
DDC
Table 4-3 Timing Requirements for OSCIN Reference Clock (14.3181818MHz)
1. Power rails from the same group are assumed to be generated by the same voltage regulator.
2. Power rails from different groups but at the same voltage can either be generated by separate regulators or by the same regulators as
long as they comply with the requirements specified in the SR5690 Motherboard Design Guide.
4.5.1 Power Up
Figure 4-1 below illustrates the power up sequencing for the various power groups, and Table 4-5
explains the symbols in the figure, as well as the associated requirements.
ACPI
STATE
VDDA18PCIE1.8VS0-S2PCI Express interface 1.8V IO and PLL
VDDA18HTPLL1.8VS0-S2HyperTransport interface 1.8V PLL power
Description
power
Figure 4-1 SR5650 Power Rail Power Up Sequence
Table 4-5 SR5650 Power Rail Power-up Sequence
SymbolParameterRequirementComment
T101.8V rails to VDDHTTX (1.2V) VDDHTTX ramps after 1.8V rails. See Note 1.
T11VDDHTTX (1.2V) to VDDPCIE (1.1V)VDDPCIE ramps together with or after VDDHTTXSee Note 1 and 2.
T12VDDHTTX(1.2V) to HT_1.1V railsHT_1.1V rails ramp together with or after VDDHTTX See Note 1 and 2.
T13VDDHTTX(1.2V) to VDDC (1.1V)VDDC ramps together with or after VDDHTTXSee Note 1 and 2.
Notes:
1. Power rail A ramps after power rail B means that the voltage of rail A does not exceed that of rail B at any time.
2. Power rail A ramps together with power rail B means that the two rails are controlled by the same enable signal and the difference
in their ramping rates is only due to the differences in the loadings.
For power down, the rails should either be turned off simultaneously or in the reversed order of the
power up sequence. Variations in speeds of decay due to different capacitor discharge rates can be
safely ignored.
Table 5-1 Power Rail Maximum and Minimum Voltage Ratings
Chapter 5
PinTypical
DC Limit*AC Limit*
UnitComments
Min.Max.Min.Max.
VDDC1.11.0671.1331.0451.155VCore power
VDD181.81.7461.8541.711.89V1.8V I/O Powers
®
VDDPCIE
VDDA18PCIE
VDDHT
VDDHTTX
VDDA18HTPLL
* Note: The voltage set-point must be contained within the DC specification in order to ensure proper operation. Voltage ripple and transient
events outside the DC specification must remain within the AC specification at all times. Transients must return to within the DC specification
within 20s.
1.11.0671.1331.0451.155V
1.81.7461.8541.711.89V
1.11.0671.1331.0451.155V
1.21.1641.2361.141.26V
1.81.7461.8541.711.89V
PCI Express
Power
PCI Express interface 1.8V I/O and
PLL power
HyperTransport™ Interface digital
I/O power
HyperTransport Transmit Interface
I/O power
HyperTransport interface 1.8V PLL
power
Interface Main I/O
Table 5-2 Power Rail Current Ratings
Min. Load
Power Rail
Average
Current
(A)
VDDC0.624.594.593.97201
VDD180.000480.000510.000510.00003-
VDDPCIE0.312.543.172.86 22
VDDA18PCIE0.020.780.910.89 16
VDDHT0.231.941.941.71 26
VDDHTTX0.080.510.510.437.3
VDDA18HTPLL0.0070.0130.0130.006-
Max. Load
Average
Current
(A)
Max. Average
Power-on
Current
(A)
Max. Step Load
Size
(A)
Max. Slew Rate
(A/µs)
5.1.2 DC Characteristics
Table 5-1 DC Characteristics for PCIe® Differential Clocks (GPP1_REFCLK and GPP3_REFCLK at 100MHz)
2) For detailed current/voltage characteristics please refer to IBIS model.
3) Measurement taken with SP/SN set to default values, PVT=Noml Case
Table 5-4 DC Characteristics for the HyperTransport™ 100MHz Differential Clock (HT_REFCLK)
SymbolDescriptionMinimumTypicalMaximumComments
V
IL
V
IH
V
IMAX
Absolute Max Input Voltage-+1.15V
Absolute Min Input Voltage--0.15V
Input High Voltage1.1-V1
Input Low Voltage-0.7V1
Minimum Output High Voltage @ I=8mA
Maximum Output Low Voltage @ I=8mA
Minimum Output Low Current @ V=0.1V
Minimum Output High Current @ V=VDDR-0.1V
Input Low Voltage–0V0.2V–
Input High Voltage1.4V1.8V––
Maximum Input Voltage––2.1V–
1.4-V2, 3
-0.4V2, 3
2.0-mA2, 3
2.0-mA2, 3
5.2SR5650 Thermal Characteristics
This section describes some key thermal parameters of the SR5650. For a detailed discussion on these parameters and
other thermal design descriptions, including package level thermal data and analysis, please consult the Thermal Design and Analysis Guidelines for SR5650/5670/5690, order# 44382.
5.2.1 SR5650 Thermal Limits
Table 5-5 SR5650 Thermal Limits
ParameterMinimumNominalMaximumUnitNote
Operating Case Temperature0—95
Absolute Rated Junction
Temperature
Storage Temperature-40—60
Ambient Temperature0—55
Thermal Design Power—12.6—W4
Notes:
1 - The maximum operating case temperature is the die top-center temperature measured via a thermocouple based on the
methodology given in the document Thermal Design and Analysis Guidelines for SR5650/5670/5690 (Chapter 12). This is the
temperature at which the functionality of the chip is qualified.
2 - The maximum absolute rated junction temperature is the junction temperature at which the device can operate without causing
damage to the ASIC.
3 - The ambient temperature is defined as the temperature of the local intake air at the inlet to the thermal management device. The
maximum ambient temperature is dependent on the heat sink design, and the value given here is based on AMD’s reference heat sink
solution for the SR5650. Refer to Chapter 6 in Thermal Design and Analysis Guidelines for SR5650/5670/5690 for heatsink and thermal
design guidelines. Refer to Chapter 7 for details of ambient conditions.
4 - Thermal Design Power (TDP) is defined as the highest power dissipated while running currently available worst case applications at
nominal voltages. The core voltage was raised to 5% above its nominal value for measuring the ASIC power. Since the core power of
modern ASICs using 65nm and smaller process technology can vary significantly, parts specifically screened for higher core power were
used for TDP measurement. The TDP is intended only as a design reference, and the value given here is preliminary.
The SR5650 has an on-die thermal diode, with its positive and negative terminals connected to the THERMALDIODE_P
and THERMALDIODE_N pins respectively. Combined with a thermal sensor circuit, the diode temperature, and hence
the ASIC junction temperature, can be derived from a differential voltage reading (V). The equation relating the
temperature to V is given below.
where:
V = Difference of two base-to-emitter voltage readings, one using current = I and the other using current = N x I
N = Ratio of the two thermal diode currents (=10 when using an ADI thermal sensor, e.g.: ADM 1020, 1030)
= Ideality factor of the diode
K = Boltzman’s Constant
T = Temperature in Kelvin
q = Electron charge
The series resistance of the thermal diode (R
every 1.0, approximately 0.8
o
C is added to the reading). The sensor circuit should be calibrated to offset the RT induced,
) must be taken into account as it introduces an error in the reading (for
T
plus any other known fixed errors. Measured values of diode ideality factor and series resistance for the diode circuit are
defined in Thermal Design and Analysis Guidelines for SR5650/5670/5690.
Note: Maximum height of SMT components is 0.650 mm.
Figure 5-3 SR5650 Ball Arrangement (Bottom View)
5.3.1 Pressure Specification
To avoid damages to the ASIC (die or solder ball joint cracks) caused by improper mechanical assembly of the cooling
device, follow the recommendations below:
•It is recommended that the maximum load that is evenly applied across the contact area between the thermal
management device and the die does not exceed 6 lbf. Note that a total load of 4-6 lbf is adequate to secure the
thermal management device and achieve the lowest thermal contact resistance with a temperature drop across
the thermal interface material of no more than 3°C. Also, the surface flatness of the metal spreader should be
0.001 inch/1 inch.
•Pre-test the assembly fixture with a strain gauge to make sure that the flexing of the final assembled board and
the pressure applying around the ASIC package will not exceed 600 micron strain under any circumstances.
•Ensure that any distortion (bow or twist) of the board after SMT and cooling device assembly is within industry
guidelines (IPC/EIA J-STD-001). For measurement method, refer to the industry approved technique described
in the manual IPC-TM-650, section 2.4.22.
5.3.2 Board Solder Reflow Process Recommendations
5.3.2.1 Stencil Opening Size for Solderball Pads on PCB
Warpage of the PCB and the package may cause solderjoint quality issues at the surface mount. Therefore, it is
recommended that the stencil opening sizes be adjusted to compensate for the warpage. The recommendation is for the
stencil aperture of the solderballs to be kept at the same size as the pads.
5.3.2.2 Reflow Profile
A reference reflow profile is given below. Please note the following when using RoHS/lead-free solder (SAC105/305/405
Tin-Silver-Cu):
•The final reflow temperature profile will depend on the type of solder paste and chemistry of flux used in the SMT
process. Modifications to the reference reflow profile may be required in order to accommodate the requirements of
the other components in the application.
•An oven with 10 heating zones or above is recommended.
•To ensure that the reflow profile meets the target specification on both sides of the board, a different profile and oven
recipe for the first and second reflow may be required.
•Mechanical stiffening can be used to minimize board warpage during reflow.
•It is suggested to decrease temperature cooling rate to minimize board warpage.
•This reflow profile applies only to RoHS/lead-free (high temperature) soldering process and it should not be used for
Eutectic solder packages. Damage may result if this condition is violated.
This chapter describes the support for ACPI power management provided by the SR5650. The SR5650 system controller
supports ACPI Revision 2.0. The hardware, system BIOS, and drivers of the SR5650 have the logic required for meeting
the power management specifications of PC2001, OnNow, and the Windows Logo Program and Device Requirements
version 2.1. Table 6-1, “ACPI States Supported by the SR5650,” describes the ACPI states supported by the SR5650
system controller.
Table 6-1 ACPI States Supported by the SR5650
ACPI StateDescription
Processor States:
S0/C0: Working StateWorking State. The processor is executing instructions.
S0/C1: Halt
S0/C2: Stop Grant
Caches Snoopable
S0/C3: Stop Grant
Caches Snoopable
System States:
S1: Standby
Powered On Suspend
S3: Standby
Suspend to RAM
S4: Hibernate
Suspend to Disk
S5: Soft OffSystem is off. OS re-boots when the system transitions to the working state.
G3: Mechanical Off
CPU Halt state. No instructions are executed. This state has the lowest latency on resume and contributes
minimum power savings.
Stop Grant or Cache Snoopable CPU state. This state offers more power savings but has a higher latency
on resume than the C1 state.
Processor is put into the Stop Grant state. Caches are still snoopable. The HyperTransport™ link may be
disconnected and put into a low power state. System memory may be put into self-refresh.
System is in Standby mode. This state has low wakeup latency on resume. OEM support of this state is
optional.
System is off but context is saved to RAM. System memory is put into self-refresh.
System is off but context is saved to disk. When the system transitions to the working state, the OS is
resumed without a system re-boot.
Occurs when system power (AC or battery) is not present or is unable to keep the system in one of the
other states.
The SR5650 system controller has integrated test modes and capabilities. These test features cover both the ASIC and
board level testing. The ASIC tests provide a very high fault coverage and low DPM (Defect Per Million) ratio of the part.
The board level tests modes can be used for motherboard manufacturing and debug purposes. The following are the test
modes of the SR5650 system controller:
•Full scan implementation on the digital core logic that provides about 97% fault coverage through ATPG
(Automatic Test Pattern Generation Vectors).
•Dedicated test logic for the on-chip custom memory macros to provide complete coverage on these modules.
•Improved access to the analog modules and PLLs in the SR5650 system controller in order to allow full
evaluation and characterization of these modules.
•A JTAG test mode (which is not entirely compliant to the IEEE 1149.1 standard) in order to allow board level
testing of neighboring devices.
•An XOR TREE test mode on all the digital I/O’s to allow for proper soldering verification at the board level.
•A VOH/VOL test mode on all digital I/O’s to allow for proper verification of output high and output low
voltages at the board level.
Chapter 7
Testability
These test modes can be accessed through the settings on the instruction register of the JTAG circuitry.
7.2Test Interface
Table 7-1 Pins on the Test Interface
Pin NameBall numberTypeDescription
TESTMODEA19ITEST_EN: Test Enable (IEEE 1149.1 test port reset)
PCIE_RESET_GPIO3D19ITMS: Test Mode Select (IEEE 1149.1 test mode select)
I2C_DATAC20ITDI: Test Mode Data In (IEEE 1149.1 data in)
I2C_CLKB20ITCLK: Test Mode Clock (IEEE 1149.1 clock)
PWM_GPIO6
PWM_GPIO4A15ITEST_ODD: Control ODD output in VOH/VOL test
PWM_GPIO3F16ITEST_EVEN: Control EVEN output in VOH/VOL test
POWERGOODA17II/O Reset
B16OTDO: Test Mode Data Out (IEEE 1149.1 data out)
7.3XOR Tree
7.3.1 Brief Description of an XOR Tree
A sample of a generic XOR tree is shown in the figure below.
Pin A is assigned to the output direction, and pins B through F are assigned to the input direction. It can be seen that after
all pins B to F are assigned to logic 0 or 1, a logic change in any one of these pins will toggle the output pin A.
The following is the truth table for the XOR tree shown in Figure 7-1 The XOR start signal is assumed to be logic 1.
The XOR start signal is applied at the TDI Pin of the JTAG circuitry and the output of the XOR tree is obtained at the
TDO Pin. Refer to
There is no specific order to these signals in the tree. A toggle of any of these balls in the XOR tree will cause the output
to toggle. When the XOR tree is activated, any pin on the XOR tree must be either pulled down or pulled up to the I/O
voltage of the pin. Only pins that are not on the XOR tree can be left floating.
When differential signal pairs are listed as single entries on the XOR tree, opposite input values should be applied to the
two signals in each pair (e.g., for entry no. 1 on the tree, when “1” is applied to HT_RXCAD0P, “0” should be applied to
HT_RXCAD0N).
Table 7-3 SR5650 XOR Tree
No.Pin NameBall Ref.
1HT_RXCAD0P/NAD28/AD27
2HT_RXCAD1P/NAC27/AC26
3HT_RXCAD2P/NAB28/AB27
4HT_RXCAD3P/NAA27/AA26
5HT_RXCAD4P/NW27/W26
6HT_RXCAD5P/NV28/V27
7HT_RXCAD6P/NU27/U26
8HT_RXCAD7P/NT28/T27
9HT_RXCTL0P/NR27/R26
10HT_RXCAD8P/NAD25/AD24
11HT_RXCAD9P/NAC24/AC23
12HT_RXCAD10P/NAB25/AB24
13HT_RXCAD11P/NAA24/AA23
14HT_RXCAD12P/NW24/W23
15HT_RXCAD13P/NV25/V24
16HT_RXCAD14P/NU24/U23
17HT_RXCAD15P/NT25/T24
18HT_RXCTL1P/NR24/R23
19GPP1_RX0P/NE11/F11
20GPP1_RX1P/ND10/E10
21GPP1_RX2P/NE9/F9
22GPP1_RX3P/ND8/E8
23GPP1_RX4P/NE7/F7
24GPP1_RX5P/ND6/E6
25GPP1_RX6P/NB5/C5
26GPP1_RX7P/ND2/D1
27GPP1_RX8P/NF5/F4
28GPP1_RX9P/NG6/G5
Table 7-3 for the list of the signals included on the XOR tree.
The VOH/VOL logic provides signal output on I/O’s when test patterns are applied to the TEST_ODD and TEST_EVEN
pins. A sample of a generic VOH/VOL tree is shown in the figure below.
7. Load JTAG instruction register with the instruction 0010 0000.
8. Load JTAG instruction register with the instruction 0101 1101.
9. Go to Run-Test_Idle state.
10. Set POWERGOOD to 1.
7.4.3 VOH/VOL pin list
Table 7-5 below shows the SR5650 VOH/VOL Tree. There is no specific order of connection. Under the Control column,
an “Odd” or “Even” indicates that the logical output of the pin is same as the input to the “TEST_ODD” or the
“TEST_EVEN” pin respectively.
When a differential signal pair appear in the table as a single entry, the output of the positive (“P”) pin is indicated in the
Control column (see last paragraph for explanations) and the output of the negative pin (“N”) will be of the opposite
value. E.g., for entry no. 1 on the tree, when TEST_EVEN is 1, HT_TXCAD0P will give a value of 1 and HT_TXCAD0N
will give a value of 0.