25325

5/27/2018 25325

1/4825325B August 2002

System Considerations for DualAMD AthlonTM MP Processors in Tower and 1U

Form Factors

5/27/2018 25325

2/48

25325B August 2002 225325B

Outline

Introduction

AMD AthlonTM MP processor power roadmap

Thermal test methodology

Tower chassis form factor

Chassis vendors and models

Power supplies

Heat sinks

Thermal test results

1U drawer form factor

Recommended 1U drawer floor plan Heat sinks

Thermal test results

Summary

5/27/2018 25325

3/48

25325B August 2002 325325B

Introduction

AMD-760TMMP chipset

Support for two AMD AthlonTM MPprocessors

Support for 200 MHz and 266 MHzfront-side bus

Support for up to 3 GB PC2100registered DDR RAM

Extended ATX Form Factor

AGP Pro slot, five 64-Bit PCI slots,VGA slot

One serial port, one parallel port, twoUSB (four through optional cable),PS2 keyboard and mouseconnections

Up to four Enhanced IDE devices

Dual 3COM LAN controllers

Integrated ATI Rage XL graphicsaccelerator

--> same motherboard used for Tower and 1U form factors

5/27/2018 25325

4/48

25325B August 2002 425325B

AMD Athlon Processor Power Roadmap

AMD Athlon MP

processor

AMD Athlon processor

40.0W

50.0W

60.0W

70.0W

80.0W

800 900 1000 1100 1200 1300 1400 1500 1600 1700 1800 1900

Frequency (MHz)

MaxTh

ermalPowe

r(W)

5/27/2018 25325

5/4825325B August 2002

Thermal Test Methodology

5/27/2018 25325

6/48

25325B August 2002 625325B

Thermal Requirements

Processor die temperature should not exceed the maximumtemperature specification (90oC or 95oC).

35oC external ambient

Maximum thermal power at specification

Northbridge, power FETs, and other components in system should

not exceed their temperature specifications.

5/27/2018 25325

7/48

25325B August 2002 725325B

Overview

Single processor SoftwareMaxtherm Option F run under DOS

Use single-processor thermal test Excel spreadsheet for analysis

Dual processor SoftwareBurnK6

Run two instances under Windows OS

Run two instances for Linux OS (under two logins)

Available at http://users.ev1.net/~redelm/ Use dual-processor thermal test Excel spreadsheet for analysis

Power dissipation: calibrate one or two processors As function of software, voltage, and temperature

Run one test with one set of processors Analyze data versus power specification

Thermocouple locations of interest

5/27/2018 25325

8/48

25325B August 2002 825325B

Test Methodology

Measurements taken at room temperature Temperature-controlled fans in power supply are run with voltage at 35C

setting.

Software used to power processor Single processor tests use Maxtherm Option

Dual processor tests utilize either BurnK6 freeware or other software

Thermocouples placed at following locations: Back of ceramic of PGA package for each processor

Heat sink base for each processor

Ambient for each processor

FET case for each processor

Under motherboard at Northbridge Power supply exhaust

System fan exhaust (if applicable)

External ambient

Vcc core feedback monitored for each processor

5/27/2018 25325

9/48

25325B August 2002 925325B

Test Methodology (cont.)

Power supply fan voltage is monitored or set at voltage at which itwould run at maximum external ambient temperature.

Determination of processor power consumption Processor placed in modified motherboard used for measuring power

Current readout measured with Vcc feedback voltage set to match voltagemeasured during thermal test

Maxtherm Option F and BurnK6 or other software run for test

Power measured at close to same temperature of test

Power used in spreadsheet = Vcc Voltage x Current (as measured throughCoreFB and CoreFB+ pins (AG11 and AG13) on the processors)

5/27/2018 25325

10/48

25325B August 2002 1025325B

Indirect Temperature Measurement Method

Simple, repeatable method

Attach thermocouple toback of ceramic substrate ofprocessor

Centered on die area

Use provided correlation fordetermining dietemperature

5/27/2018 25325

11/48

25325B August 2002 1125325B

Correlation for Indirect Die TemperatureMethod for Products without Diodes

Correlation relates dietemperature to backsideceramic temperature

Referenced to local CPUambient

Measurementsperformed on a thermal

test vehicle

Power varied to createcorrelation

Validated againstmultiple, more involvedmethods of makingtemperaturemeasurement

Typical residual: 2C for centered clip

2(TC3-Tamb) = 1.209x - 1.3778

R2

= 0.9998

20

25

30

35

40

45

50

55

60

65

70

20 25 30 35 40 45 50 55 60 65 70

(T ceramic T CPU ambient) (C)

(Die

TempTC

PUambien

t)(C)

C3

x = y

Linear (C3)

5/27/2018 25325

12/48

25325B August 2002 1225325B

CPU Ambient Thermocouple

Thermocouple should becentered above the fan hub

One inch above the center of

the fan hub

Tape it to stiff wire that can beformed to place it in thislocation

5/27/2018 25325

13/48

25325B August 2002 1325325B

Evaluation Temperature Sensor Kits

Maxim 1617A Evaluation Kit

Analog 1021A Evaluation Kit

5/27/2018 25325

14/48

25325B August 2002 1425325B

Connect to the On-Die Temperature Diode

Attach temperature sense wires to the evaluation kit S7 wire connects to the positive terminal (DXP1 or D+)

U7 wire connects to the negative terminal (DXN1 or D)

Remove

this

remotediode

Positive

terminal

Negativeterminal

5/27/2018 25325

15/48

25325B August 2002 1525325B

Connect to Vcc Core Feedback Pins

Connect 30-gage wire to Vcc core feedback positive and negativepins AG11 and AG13 (see processor pinout on next page)

Allows monitoring of exact voltage across processor

5/27/2018 25325

16/48

25325B August 2002 1625325B

Processor System Data Analysis

Room temperature data extrapolated to 35C

Die temperature calculated using following correlation for ceramicpackages:

(Tdie Tamb) = 1.209 (Tceramic Tamb) 1.3778

!js and ! sa calculated

Maximum Allowable Power (MAP) at 35C calculated:

MAP = (90 C or 95CTamb)/ !ja

Projected die temperature calculated:

Tdie = Power(from MTP table) !ja+Tamb

5/27/2018 25325

17/48

25325B August 2002 1725325B

Data Analysis

MTP Table: MAP compared to Max Thermal Power(MTP)

Example: MAP = 54W

From table, maximum frequencysupport = 1000MHz

Max thermal power represents +3sigma power dissipation of

product

at a given frequency and nominalvoltage

650 36.1

700 38.3

750 40.5

800 42.6

850 44.8900 49.1

950 51.4

1000 51.0

1100 55.3

1200 59.6

1300 63.91400 68.2

AMD AthlonTM

Processor1.75 V SpecFP*

frequency Max Thermal power:

D/C power 9

54 W maxthermal power

surpasses

1.0 GHz, but is

not enough for

1.1 GHz.

5/27/2018 25325

18/48

25325B August 2002

Tower Form Factor Guidelines

5/27/2018 25325

19/48

25325B August 2002 1925325B

Thermal Design for Dual AMD Athlon MPSystems in Tower Form Factor

Layout motherboard using keepout region definition for singleprocessor systems

Use a single qualified heat sink for each processor Use system fans to reduce temperature rise from external to the

chassis to the processor locations

Manage acoustics of system through fan speed control

Chassis are standard product offerings from various chassis vendors

5/27/2018 25325

20/48

25325B August 2002 2025325B

Chassis

ChenMing 601AE-F-D *

Palo Alto 810*

Chenbro 2012 / 2025* Evercase EC S5000

Inwin QS5000

* will present test results*Will present test results

5/27/2018 25325

21/48

25325B August 2002 2125325B

Heat Sinks for Dual ProcessorImplementations in Tower Form Factor

Use standard single processor desktop heat sinks.

Heat sink frequency rating is based on the following criteria: Maximum external system ambient = 35C

External-to-internal temperature rise " 14C

Power dissipation is at the maximum specification.

Die temperature " maximum specification

External-to-internal temperature rise must meet the above criteria.

IMPORTANT: System integrator must confirm that processortemperature specification is met.

5/27/2018 25325

22/48

25325B August 2002 2225325B

Heat Sink Incorporating Copper

Foxconn PK0453AEDAU52

63 x 70 x 60 mm

Basic aluminum extrusion

50 x 50 x 3mm Ni-plated Cu slug

Attached with four screws

Shinetsu G751 grease

5/27/2018 25325

23/48

25325B August 2002

Dual AMD Athlon Processor Tower

System Integration Example

5/27/2018 25325

24/48

25325B August 2002 2425325B

Tower Chassis2P System Configuration

System configuration

Processors: 1400 Mhz AMD AthlonTM MP processors

Motherboard: Tyan Thunder

Power supply: Delta 460W

SCSI drives: Quantum Atlas 9GB

Video card: Elsa Gloria II

Memory: 1024MB Samsung PC2100

DVD drive: Toshiba SD-M1402 Floppy disk drive: TEAC FD-235HF

Sound card: Sound Blaster Live

Tower chassis: ChenMing, Chenbro or Palo Alto

5/27/2018 25325

25/48

25325B August 2002 2525325B

System Configuration TestedChenMing

5/27/2018 25325

26/48

25325B August 2002 2625325B

Measurements from ChenMing Test

0

10

20

30

40

50

60

70

80

90

5 10 15 20 25 30

Time (min)

Temper

ature(C)

Die 0

Heatsink Base 0

CPU Amb. 0

FET 0

Die 1

Heatsink Base 1

CPU Ambient 1

FET 1

SCSI Drive

DIMM

NB

External Ambient

Temperatu

re(C)

5/27/2018 25325

27/48

25325B August 2002 2725325B

Thermal Characterization Results

Test Conditions Dual processor mode running two instances of BurnK61400 MHz

AMD AthlonTM MP processors

Foxconn PK0453AEDAUFB aluminum/copper heatsinks utilized Optimal performance obtained with one 80-mm system fan located in front of

chassis

ChenMing Results

Supports dual AMD AthlonTM MP processor-based 1200-MHz system with aboveconfiguration

Rear processor --> processor 1 is always highest temperature

Further optimization required to support beyond 1200 MHz

IR Image of Mothe boa d in

5/27/2018 25325

28/48

25325B August 2002 2825325B

IR Image of Motherboard inChenbro Chassis

5/27/2018 25325

29/48

25325B August 2002 2925325B

Chenbro Thermal Characterization

Test conditions

Dual processor mode running two instances of BurnK61400-MHzAMD AthlonTM MP processors

Foxconn PK0453AEDAUFB aluminum/copper heatsinks utilized

Optimal performance obtained with one 80-mm front system fan and one80-mm rear system fan

Chenbro Results

Supports dual AMD AthlonTM MP processor 1000-MHz system with aboveconfiguration



5/27/2018 25325

30/48

25325B August 2002 3025325B

Palo Alto Thermal Characterization

Test conditions

Dual processor mode running two instances of BurnK61400-MHzAMD AthlonTM MP processors

Foxconn PK0453AEDAUFB aluminum/copper heatsinks utilized

Optimal performance obtained with one 90-mm system fan in front andone 120-mm fan in rear of chassis

Palo Alto Results

Support a dual 1200-MHz system



5/27/2018 25325

31/48

25325B August 2002 3125325B

Future Development Focus

Efforts to improve frequency support:

Ducting

Heatsinks with different fin orientation

Efforts to improve heatsink testing:

Speed controlled fans to reduce noise at room temperature

Ducting to eliminate need for system fans

5/27/2018 25325

32/48

25325B August 2002

Recommendations for DualAMD AthlonTM MP Processor-Based Systems

in 1U Rack Mounted Form Factor

Thermal Design Approach for Dual AMD Athlon MP

5/27/2018 25325

33/48

25325B August 2002 3325325B

g ppProcessor-Based Systems in 1U Form Factor

Layout motherboard utilizing keepout region definition for singleprocessor systems

Drawer floorplan is custom for the motherboard. Recommendedfloorplan is described using Tyan Thunder motherboard

Utilize a single heat sink per processor

Heat sink is coupled to blower placed in close proximity to the board

Utilizes low profile extrusions used for desktop without dedicated fan on heatsinks

Power supplies specification in review with development partners

Reference Design 1U Power Supply

5/27/2018 25325

34/48

25325B August 2002 3425325B

Reference Design 1U Power SupplyRequirements

Output Tolerance Max Current , Amps3.3 VDC +/ 5% 8 A

5 VDC +/ 5% 15 A

12 VCPU +/ 5% 16 A

12 VIO +/ 5% 5A (7A Peak)

12 VDC +/10% 0.8 A

5 VR +/ 5% 2.0 A

> 8 cfm at low speed

>10 cfm at high speed

Heat Sinks for Dual Processor

5/27/2018 25325

35/48

25325B August 2002 3525325B

Heat Sinks for Dual ProcessorImplementations in 1U Form Factor

Address thermal design at the drawer layout/floor plan level Low-profile fan sinks do not have sufficient pressure/flow characteristics for this

form factor.

Standard 1U designs that have a simple front-to-back flow distribution do notprovide adequate cooling.

Custom floor plan utilizing higher pressure/flow blowers for cooling the processoris required to support entire roadmap

IMPORTANT: Entire floor plan evaluated to assess performance Ducting and blowers

Cooling provided to drives, memory, and power supply

Temperature specification guidelines Maximum external system ambient = 35C

Power dissipation is at the maximum specification

Die temperature " maximum specification

Example Heat Sink Used in

5/27/2018 25325

36/48

25325B August 2002 3625325B

pReference Design

Base: 60mm width x 60mm depth x 7.5mm thickness

Cu slug: 45mm width x 60mm depth x 4mm thickness

Fin: 22.6mm height x 0.92mm width x 2.04mm spacing

Crosscuts: 3.05mm wide

Thermal interface material: Thermagon TCP 905c

Example 97-mm Blowers Used in 1U

5/27/2018 25325

37/48

25325B August 2002 3725325B

pReference Design

Nidec Model #A34124-16

0.65A @ 12V

23 cfm at 0 H2O

0.95 of H2O at 0 cfm

Delta Model# BFB1012 H

1.2A @ 12V

25 cfm at 0 H2O

1.0 of H2O at 0 cfm

Both blowers have been used with comparable results.

Matching Air Movers to System Resistance

5/27/2018 25325

38/48

25325B August 2002 3825325B

Matching Air Movers to System Resistance

System curvefor old ducting

Two 97-mm blowersin parallel

Four fans in parallel

in series with twofans in parallel

5/27/2018 25325

39/48

25325B August 2002

1U System Integration Example

1U System Configuration

5/27/2018 25325

40/48

25325B August 2002 4025325B

1U System Configuration

System Configuration Processors: 1200 MHz AMD AthlonTM MP

Motherboard: Tyan Thunder

Power supply: In development

SCSI drives: Quantum Atlas 10KII 18.4B

Onboard video card: ATI Rage 128

Memory: 1024Mb Samsung PC2100

CD drive: Toshiba XM-1902B

Floppy disk drive: TEAC FD-235HF

1U server chassis: Computer and Control Solutions

AMD Proposed Configuration

5/27/2018 25325

41/48

25325B August 2002 4125325B

AMD Proposed Configuration

SCA

Drive

SCA Drive

CD ROM

& Floppy

DiscPower

SupplyTyan Thunder

CPU 0CPU

197mm Blower

97mm Blower

Ducting not shown on processor

Front

RearPlace power supply to center of drawer to

allow placement of 97-mm blowers

adjacent to processors.

Utilize drawer length of 2728.

Duct flow from blowers directly to

processor locations.

Reference Floor Plan in Lexan Chassis

5/27/2018 25325

42/48

25325B August 2002 4225325B

Reference Floor Plan in Lexan Chassis

Three Blowers* Power Supply

Ducted Heat Sinks Yellow Line = Seal Against Flow

Primary Exhaust

*Reduced to two blowers in the final configuration.

Temperature Measurements in 1U Floorplan

5/27/2018 25325

43/48

25325B August 2002 4325325B

Temperature Measurements in 1U Floorplan

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

100.0

0 10 20 30 40 50 60

Time (minutes)

T

emperature(C)

Processor 0

DDR DIMMCPU Amb.0

FET0

Processor 1

FET1

CPU Amb.1

SCSI DriveNorthbridge

Amb.

Te

mperature(C)

Measured using 800-MHz AMD AthlonTM processors

3C per 100MHz --> ~1000 MHz of margin to 95C specification

Test Data for Floor Plan Proposed by Floor Plan AMD

5/27/2018 25325

44/48

25325B August 2002 4425325B

Test Data for Floor Plan Proposed by Floor Plan AMD

CPU 0 CPU1

ConfigurationExternalTemp. JA C/W

MHz at35C amb

ExternalTemp. JA

C/W MHz at35C amb

Lexan chassis, dual 800 AMD AthlonTM

processors, Taisol 20 Fin 60x60 heatsinks,

three NMB blowers, dual 18G SCA drives4.3 0.666 1800 4.2 0.806 1500

Lexan chassis, dual 800 AMD AthlonTM

processors, Taisol 20 Fin 60x60 heatsinks,

two NMB blowers, dual 18G SCA drives6.5 0.595 2000 5.6 0.762 1600

Floor plan covers up to 1600-MHz AMD Athlon MP processors

Optimization being performed to push the performance higher


5/27/2018 25325

45/48

25325B August 2002 4525325B


Blower selection

Optimization of pressure drops in ducting

Optimizing heat sinksSelect fin density to match blowercharacteristics.

Optimize the air flow path through the drawer to reduce the numberof turns, expansions, and contractions.

Summary

5/27/2018 25325

46/48

25325B August 2002 4625325B

Summary

2P Tower form factor Three systems tested and validated

ChenbroQualified through 1000 MHz

ChenMingQualified through 1200 MHz

Palo AltoQualified through 1200 MHz

Further work ongoing to improve frequency coverage of above systems

1U form factor Reference design developed

Working with system and chassis suppliers for enabled solutions

2001, 2002 Advanced Micro Devices, Inc.

5/27/2018 25325

47/48

25325B August 2002 4725325B

The contents of this document are provided in connection withAdvanced Micro Devices, Inc. (AMD) products. AMD makes norepresentations or warranties with respect to the accuracy or

completeness of the contents of this publication and reserves the rightto make changes to specifications and product descriptions at any timewithout notice. No license, whether express, implied, arising byestoppel or otherwise, to any intellectual property rights is granted by

this publication. Except as set forth in AMDs Standard Terms andConditions of Sale, AMD assumes no liability whatsoever, and disclaimsany express or implied warranty, relating to its products including, butnot limited to, the implied warranty of merchantability, fitness for aparticular purpose, or infringement of any intellectual property right.

5/27/2018 25325

48/48

25325B August 2002 4825325B

AMDs products are not designed, intended, authorized or warrantedfor use as components in systems intended for surgical implant into thebody, or in other applications intended to support or sustain life, or in

any other application in which the failure of AMDs product could createa situation where personal injury, death, or severe property orenvironmental damage may occur. AMD reserves the right todiscontinue or make changes to its products at any time without

notice.

AMD, the AMD Arrow logo, AMD Athlon, and combinations thereof andAMD-760 are trademarks of Advanced Micro Devices, Inc. Windows is

a registered trademark of Microsoft corporation. Other product namesused in this publication are for identification purposes only and may betrademarks of their respective companies.

25325

Documents