Sustainable and Secure - It's EPYC - AMD.pdf

1、Lynn Comp2|Public3|PublicEndnote:SP5-013B4|Public10 Containers of8 cores each24 Containers of8 cores each2x 4th GenAMD EPYC2x 3rd GenXeon Platinum10 Containers of8 cores each24 Containers of8 cores each2x 4th GenAMD EPYC2x 3rd GenXeon PlatinumEndnotes:SP5-070 and-071A5|PublicEndnote:SP5TCO-019B6|Pub

2、licIf was measuredin acres of forest Endnote:SP5TCO-021A133m VMs based on VMmark tiles.7|Public2022 Update on 30 x25 Energy Efficiency Goal Ahead of the Curvehttps:/ Endnotes:Footnote 18|PublicNon-AMD AWS EC2 instance32c/300WAWS M6a(AMD)Milan 7R13/48c/225WCPU power per server600W(2x300w)450W(2x225W)

3、DDR4 power per server128W(512GB)141W(768GB)Miscellaneous power per server(Board,PS,NIC,SSD,etc.)75W75WTotal power per server803W666WTotal#servers per rack 65%utilization(power limit)3340Total#servers per rack(physical limit 1.5U per server)2626Total#of vCPUs per rack3328(26x2x2x32)4992(26x2x2x48)Tot

4、al rack power(65%utilization)13,570W(803x0.65x26)11,255W(666x0.65x26)Rack Energy Efficiency(W/vCPU)Lower is Better4.12.3Rack Energy Efficiency(W/vCPU)AMD EPYC Milan based EC2 VMs are 44%more energy efficient vs.Non-AMD EC2 based VMs42u,17.5kVA Rack,1.5u Compute ServerAMD EPYC CPUs Make Cloud Infrast

5、ructure 44%More Energy Efficient9|Public32,66736,38838,00931,86221,06522,4968,59610,84705,00010,00015,00020,00025,00030,00035,00040,00010%20%30%40%50%60%70%80%Performanceto Power RatioTarget Power LoadSPECpower_ssj2008 ssj_ops/watt target operational loadEPYC 9654;1NEPYC 7763;1NIntel 8480+;1NAmpere

6、Q64-30;2NOperations/Watt=Work accomplished for power used1n 2P 3rdGen EPYC 776325,302 overall ssj_ops/W1n 2P 4th Gen EPYC 965430,602 overall ssj_ops/W1n 2P Xeon Platinum 8480+16,828 overall ssj_ops/W2n 1P Ampere Altra Q64-308,436 overall ssj_ops/WEnergy Consumption*=The production of a certain amoun

7、t of power for a certain amount of timeComparison based on best overall result for the named systems/CPU as published at www.spec.org as of 3/23/2023.SPEC and the benchmark name SPECpower_ssj are registered trademarks of the Standard Performance Evaluation Corporation.For more information about SPEC

8、power,see www.spec.org/power_ssj2008/.1-node 2P EPYC 9654 https:/spec.org/power_ssj2008/results/res2022q4/power_ssj2008-20221204-01204.html;1-node 2P EPYC 7763 https:/spec.org/power_ssj2008/results/res2022q3/power_ssj2008-20220617-01179.html;1-node 2P Xeon Platinum 8480+https:/spec.org/power_ssj2008

9、/results/res2023q1/power_ssj2008-20221207-01216.html;2-node 1P Ampere Altra Q64-30 https:/spec.org/power_ssj2008/results/res2022q4/power_ssj2008-20221113-01196.html.*https:/www.electricaltechnology.org/2020/12/energy-power-consumption-kwh-calculator.htmlGenoa 3.8x betterGenoa 58%better10|PublicAMD E

10、PYC CPUs Help Businesses Meet Sustainability Goals“Requests per Watt is one of our defining characteristics.We evaluated Intels latest generation of“Ice Lake”Xeon processors.Although Intels chips were able to compete with AMD in terms of raw performance,the power consumption was several hundred watt

11、s higher per server-thats enormous.This meant that Intels Performance per Watt was unattractive”Cloudflare Blog by Chris Howells“With AMD EPYC processors,we are seeing 15%more performance while helping to reduce our carbon footprint and our IT operating costs have come down by 10%.”Niko Hossain,CEO

12、of Lobster Logistics Cloud“Through the latest generation of AMD EPYC processors and Nokias cloud-native Core software,we are helping CSPs shrink the carbon footprint of their networks.This is critical as advanced 5G service roll-out accelerates,with the associated implications for new demands on ene

13、rgy consumption”Fran Heeran,SVP Core Networks at Nokiahttps:/ on increase in security keys in 4thGen EPYC CPUs of 1009 versus 3rdGen EPYC with 509 keys.*Endnote:GD-18312|Public00AMD SoCAMD Secure ProcessorRoot of TrustEncrypt data in use Encrypt data while its being processed.Help isolate it from ma

14、licious users,the hypervisor,even admins.Migrate easily Efficiently move current x86 instances to AMD EPYCpowered instances.No application changes required.Dont compromise when leveraging sophisticated security featuresFinancial ServicesHelp mitigate the risk of disclosure or alteration of financial

15、 data.HealthcareHelp protect patient records against unauthorizedaccess.Retail Help secure customer data,including transactions.Manufacturing Automate confidently across on-premises and cloud infrastructures.13|PublicAMD Infinity Guard In ActionHealthCareFinancialsTelecommunicationshttps:/ cost bene

16、fitswe realize*allow us totest and support moreworkloads for newproducts and keepeverything movingahead within budget.“Vidur Apparao,SeniorVice President ofEngineering,Salesforce*When using Google N2D VMs powered by AMDhttps:/ of August 2022,select Ryzen 7040 processors for mobile with dedicated AI

17、hardware are the only x86 PC processors with dedicated AI hardware.PHX-3Footnote 1:Includes high-performance CPU and GPU accelerators used for AI training and High-Performance Computing in a 4-Accelerator,CPU hosted configuration.Goal calculations are based on performance scores as measured by stand

18、ard performance metrics(HPC:Linpack DGEMM kernel FLOPS with 4k matrix size.AI training:lower precision training-focused floating-point math GEMM kernels such as FP16 or BF16 FLOPS operating on 4k matrices)divided by the rated power consumption of a representative accelerated compute node including t

19、he CPU host+memory,and 4 GPU accelerators.EPYC-038:Based on AMD internal testing as of 09/19/2022,geomean performance improvement at the same fixed-frequency on a 4th Gen AMD EPYC 9554 CPU compared to a 3rd Gen AMD EPYC 7763 CPU using a select set of workloads(33)including est.SPECrate2017_int_base,

20、est.SPECrate2017_fp_base,and representative server workloads.SPEC and SPECrate are registered trademarks of Standard Performance Evaluation Corporation.Learn more at spec.org.EPYC-040:AMD EPYC 9004 CPUs support 12 channels of up to 4800 MHz DDR5 memory which is 460.8 GB/s of maximum memory throughpu

21、t per socket.Prior generation of EPYC CPUs(7003 series)have a maximum 204.8 GB/s.EPYC 9004 CPUs have 2.25x the memory throughput per CPU.460.8 204.8=2.3x(2.25x)the max memory throughput.EPYC-041:40%less area measures Core+L2 Area:“Zen 4”=3.84 mm2 vs.“Sunny Cove”6.5mm2.48%estimated SPECrate2017_int_b

22、ase GCC-O3 Jemalloc results based on internal AMD reference platform and Intel platform measurements of 11/10/2022.Comparison of estimated 1P AMD EPYC 9534(537 est.SPECrate2017_int_base,set to 270 Total TDP W,64 Total Cores,AMD Est)is 1.48x the performance per watt of 1P Intel Xeon Platinum 8380(363

23、 est.SPECrate2017_int_base,270 Total TDP W,40 Total Cores,AMD est.)for 1.48x the SoC performance/watt.SPEC,SPEC CPU,and SPECrate are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more information.OEM published scores will vary based on system configura

24、tion and determinism mode used(claim uses 270W cTDP performance profile).SP5-009C:SPECrate2017_fp_base based on published scores from www.spec.org as of 11/10/2022.Configurations:2P AMD EPYC 9654(1480 SPECrate2017_fp_base,192 total cores,www.spec.org/cpu2017/results/res2022q4/cpu2017-20221024-32605.

25、html)is 2.52x the performance of published 2P Intel Xeon Platinum 8380(587 SPECrate2017_fp_base,160 total cores,www.spec.org/cpu2017/results/res2022q4/cpu2017-20221010-32542.html).Published 2P AMD EPYC 7763(663 SPECrate2017_fp_base,128 Total Cores,http:/spec.org/cpu2017/results/res2021q4/cpu2017-202

26、11121-30146.html)is shown at 1.13x for reference.SPEC,SPEC CPU,and SPECrate are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more information.SP5-010B:SPECrate2017_int_base based on published scores from www.spec.org as of 11/10/2022.Configurations:2P

27、 AMD EPYC 9654(1790 SPECrate2017_int_base,192 total cores,www.spec.org/cpu2017/results/res2022q4/cpu2017-20221024-32607.html)is 2.97x the performance of published 2P Intel Xeon Platinum 8380(602 SPECrate2017_int_base,80 total cores,http:/spec.org/cpu2017/results/res2021q2/cpu2017-20210521-26364.html

28、).Published 2P AMD EPYC 7763(861 SPECrate2017_int_base,128 total cores,http:/spec.org/cpu2017/results/res2021q4/cpu2017-20211121-30148.html)is shown at 1.43x for reference.SPEC,SPEC CPU,and SPECrate are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for mor

29、e information.SP5-012B:SPECjbb 2015-MultiJVM Max based on published scores from www.spec.org as of 11/10/2022.Configurations:2P AMD EPYC 9654(815459 SPECjbb2015 MultiJVM max-jOPS,356204 SPECjbb2015 MultiJVM critical-jOPS,192 Total Cores,http:/www.spec.org/jbb2015/results/res2022q4/jbb2015-20221019-0

30、0861.html)is 2.85x the performance of published 2P Intel Xeon Platinum 8380(286125 SPECjbb2015 MultiJVM max-jOPS,152057 SPECjbb2015 MultiJVM critical-jOPS,80 Total Cores,http:/www.spec.org/jbb2015/results/res2021q4/jbb2015-20211006-00706.html).2P AMD EPYC 7763(420774 SPECjbb2015 MultiJVM max-jOPS,16

31、5211 SPECjbb2015 MultiJVM critical-jOPS,128 total cores,http:/www.spec.org/jbb2015/results/res2021q3/jbb2015-20210701-00692.html)shown at 1.47x for reference.SPEC and SPECjbb are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more information.SP5-013A:9

32、6-core EPYC 9654 CPU processors results as of 11/10/2022 using SPECrate2017_int_base.The AMD EPYC scored 1790 SPECrate2017_int_base which is higher than all other 2P scores published on the SPEC website.2P AMD EPYC 9654(1790 SPECrate2017_int_base,192 total cores,www.spec.org/cpu2017/results/res2022q

33、4/cpu2017-20221024-32607.html).SPEC,SPECrate and SPEC CPU are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more information.SP5-016A:SPECrate2017_int_base comparison is based on a compliant ASUSTeK run and published scores from www.spec.org as of 11/1

34、0/2022.Comparison of compliant 2P AMD EPYC 9374F(815 SPECrate2017_int_base,64 Total Cores,compliant run ASUSTeK RS700A-E12,1536 GB-24x 64 GB 2Rx4 PC5-4800B-R,SUSE Linux Enterprise Server 15 SP4,AOCC 4.0)is 1.55x the performance of published 2P Intel Xeon Platinum 8362(526 SPECrate2017_int_base,64 To

35、tal Cores,http:/spec.org/cpu2017/results/res2021q3/cpu2017-20210802-28469.html).Published 2P AMD EPYC 75F3(596 SPECrate2017_int_base,64 Total Cores,http:/spec.org/cpu2017/results/res2021q2/cpu2017-20210409-25541.html)is shown for reference.SPEC,SPEC CPU,and SPECrate are registered trademarks of the

36、Standard Performance Evaluation Corporation.See www.spec.org for more information.SP5-049A:VMmark 3.1.1 matched pair comparison based on published results as of 11/10/2022.Configurations:2-node,2P 96-core EPYC 9654 powered server running VMware ESXi 8 RTM(40.19 44 tiles/836 VMs,https:/ 2-node,2P 40-

37、core Xeon Platinum 8380 running VMware ESXi v7 U2 (14.19 14 tiles/266 VMs,https:/ 2.8x the score and 3.1x the tile(VM)capacity.2-node,2P EPYC 7763-powered server(23.33 24 tiles/456 VMs,https:/ at 1.6x the performance for reference.VMmark is a registered trademark of VMware in the US or other countri

38、es.18|PublicEndnotesSP5-065:SPECrate2017_int_energy_base comparison based on published results as of 11/10/2022.Configurations:2P AMD EPYC 9654(1890 SPECrate2017_int_energy_base/1190 SPECrate2017_int_base,192 total cores,www.spec.org/cpu2017/results/res2022q4/cpu2017-20221024-32633.html)vs.2P Intel

39、Xeon Platinum 8380(725 SPECrate2017_int_energy_base/531 SPECrate2017_int_base,80 total cores,www.spec.org/cpu2017/results/res2021q2/cpu2017-20210412-25603.html).2P AMD EPYC 7713(1610 SPECrate2017_int_energy_base/576 SPECrate2017_int_base,128 total cores,www.spec.org/cpu2017/results/res2021q1/cpu2017

40、-20210301-25148.html)shown at 2.22x for reference.SPEC and SPECrate are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more information.NOTE:Red text only needs to be included with charts that show the 7763.SP5-068:SPECrate2017_int_base,SPECrate2017_fp_

41、base,and BERT-large estimates based on internal AMD reference platform measurements of 11/3/2022.Floating-point throughput comparison:2P AMD EPYC 9534(1030 est.SPECrate2017_fp_base,560 Total TDP W,128 Total Cores)is 1.66x the performance/W of published 2P AMD EPYC 7763(622 est.SPECrate2017_fp_base,5

42、60 Total TDP W,128 Total Cores).Integer throughput comparison:2P AMD EPYC 9534(1070 est.SPECrate2017_int_base,560 Total TDP W,128 Total Cores)is 1.34x the performance/W of published 2P AMD EPYC 7763(800 est.SPECrate2017_int_base,560 Total TDP W,128 Total Cores).Bert-Large NLP sparse INT8 comparison:

43、2P AMD EPYC 9534(345.6 items/sec,560 Total TDP W,128 Total Cores)is 2.67x the performance/W of published 2P AMD EPYC 7763(129.7 items/sec,560 Total TDP W,128 Total Cores).SPEC,SPEC CPU,and SPECrate are registered trademarks of the Standard Performance Evaluation Corporation.See www.spec.org for more

44、 information.OEM published scores will vary based on system configuration and determinism mode used(default cTDP performance profile).SP5-070:MySQL 8.0.17 DSS comparison based on AMD measured median scores on 2P 96-core EPYC 9654 compared to 2P 40-core Xeon Platinum 8380 running virtualized HammerDB

45、 TPROC-H SF1(KVM Hypervisor Virtualization server environment with 4 streams,4 virtual units,calculating throughput with 4 streams x 22 queries x 3600 divided by the slowest VU completion time in seconds)as of 11/10/2022.Configurations:2x AMD EPYC 9654(126,980 TPROC-H tpm)vs.2x Xeon Platinum 8380(47

46、452 TPROC-H queries/hour)for 2.68x the tpm performance.SP5-071:MySQL 8.0.17 OLTP comparison based on AMD measured median scores on 2P 96-core EPYC 9654 compared to 2P 40-core Xeon Platinum 8380 running virtualized HammerDB TPROC-C(KVM Hypervisor Virtualization server environment with 400 WH and 64 u

47、sers)as of 11/10/2022.Configurations:2x AMD EPYC 9654(126,980 TPROC-C tpm/531,183 NOPM)vs.2x Xeon Platinum 8380(47452 TPROC-C tpm/224,126 NOPM)for 2.37x the tpm/NOPM performance.SP5TCO-019B:As of 11/15/2022 based on AMD Internal analysis using the AMD EPYC Bare Metal Server&Greenhouse Gas Emission T

48、CO Estimation Tool-version 6.40,estimating the cost and quantity of 2P AMD EPYC 9654(96 core/CPU)powered server versus 2P Intel Xeon Gold 8380(40 core/CPU)based server solutions required to deliver 1995 total virtual machines(VM)based on VMmark tiles in published results.Environmental impact estimat

49、es made leveraging this data,using the Country/Region specific electricity factors from the 2020 Grid Electricity Emissions Factors v1.4 September 2020,and the United States Environmental Protection Agency Greenhouse Gas Equivalencies Calculator.This scenario contains many assumptions and estimates

50、and,while based on AMD internal research and best approximations,should be considered an example for information purposes only,and not used as a basis for decision making over actual testing.SP5TCO-021A:This analysis is a theoretical exercise,made as of 11/15/2022 based on AMD Internal analysis usin

51、g the AMD EPYC Bare Metal Server&Greenhouse Gas Emission TCO Estimation Tool-version 6.40 estimating the cost and quantity of 2P AMD EPYC 9654(96 core/CPU)powered server versus 2P Intel Xeon Gold 8380(40 core/CPU)based server solutions required to deliver 133 million total virtual machines(VM)based

52、on VMmark tiles in published results.Environmental impact estimates made leveraging this data,using the Country/Region specific electricity factors from the 2020 Grid Electricity Emissions Factors v1.4 September 2020,and the United States Environmental Protection Agency Greenhouse Gas Equivalencies

53、Calculator.This scenario contains many assumptions and estimates and,while based on AMD internal research and best approximations,should be considered an example for information purposes only,and not used as a basis for decision making over actual testing.For additional details,see https:/ of 01/18/

54、2023 based on AMD Internal analysis using the AMD EPYC Server Virtualization&Greenhouse Gas Emission TCO Estimation Tool-version 12.10 estimating the cost and quantity of 2P AMD 96 core EPYC 9654 powered server versus 2P Intel Xeon 60 core Platinum 8490H based server solutions required to deliver 20

55、00 total virtual machines(VM),requiring 1 core and 8GB of memory per VM for a 3-year period.This includes VMware software license cost of$6,558.32 per socket+one additional software for every 32 CPU core increment in that socket.Environmental impact estimates made leveraging this data,using the Coun

56、try/Region specific electricity factors from the 2020 Grid Electricity Emissions Factors v1.4 September 2020,and the United States Environmental Protection Agency Greenhouse Gas Equivalencies Calculator.This scenario contains many assumptions and estimates and,while based on AMD internal research an

57、d best approximations,should be considered an example for information purposes only,and not used as a basis for decision making over actual testing.For additional details,see https:/ PACKAGING2023 Advanced Micro Devices,Inc.All rights reserved.AMD,the AMD Arrow logo,Alveo,EPYC,RDNA,Ryzen,Instinct,Pe

58、nsando,Threadripper,Xilinx and combinations thereof are trademarks of Advanced Micro Devices,Inc.Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies.The information presented in this document is for informational purpo

59、ses only and may contain technical inaccuracies,omissions,and typographical errors.The information contained herein is subject to change and may be rendered inaccurate for many reasons,including but not limited to product and roadmap changes,component and motherboard version changes,new model and/or

60、 product releases,product differences between differing manufacturers,software changes,BIOS flashes,firmware upgrades,or the like.Any computer system has risks of security vulnerabilities that cannot be completely prevented or mitigated.AMD assumes no obligation to update or otherwise correct or rev

61、ise this information.However,AMD reserves the right to revise this information and to make changes from time to time to the content hereof without obligation of AMD to notify any person of such revisions or changes.This information is provided as is.”AMD makes no representations or warranties with r

62、espect to the contents hereof and assumes no responsibility for any inaccuracies,errors,or omissions that may appear in this information.AMD specifically disclaims any implied warranties of non-infringement,merchantability,or fitness for any particular purpose.In no event will AMD be liable to any person for any reliance,direct,indirect,special,or other consequential damages arising from the use of any information contained herein,even if AMD is expressly advised of the possibility of such damages.