- 5th Gen EPYC CPU Enterprise and Cloud Server Workloads generational IPC Uplift of 1.170x (geomean) using a select set of 36 workloads and is the geomean of estimated scores for total and all subsets of SPECrate®2017_int_base (geomean ), estimated scores for total and all subsets of SPECrate®2017_fp_base (geomean), scores for Server Side Java multi instance max ops/sec, representative Cloud Server workloads (geomean), and representative Enterprise server workloads (geomean).
“Genoa” Config (all NPS1): EPYC 9654 BIOS TQZ1005D 12c12t (1c1t/CCD in 12+1), FF 3GHz, 12x DDR5-4800 (2Rx4 64GB), 32Gbps xGMI;
“Turin” config (all NPS1): EPYC 9V45 BIOS RVOT1000F 12c12t (1c1t/CCD in 12+1), FF 3GHz, 12x DDR5-6000 (2Rx4 64GB), 32Gbps xGMI
Utilizing Performance Determinism and the Performance governor on Ubuntu® 22.04 w/ 6.8.0-40-generic kernel OS for all workloads.
- 5th Gen EPYC generational ML/HPC Server Workloads IPC Uplift of 1.369x (geomean) using a select set of 24 workloads and is the geomean of representative ML Server Workloads (geomean), and representative HPC Server Workloads (geomean).
“Genoa” Config (all NPS1) “Genoa” config: EPYC 9654 BIOS TQZ1005D 12c12t (1c1t/CCD in 12+1), FF 3GHz, 12x DDR5-4800 (2Rx4 64GB), 32Gbps xGMI;
“Turin” config (all NPS1): EPYC 9V45 BIOS RVOT1000F 12c12t (1c1t/CCD in 12+1), FF 3GHz, 12x DDR5-6000 (2Rx4 64GB), 32Gbps xGMI
Utilizing Performance Determinism and the Performance governor on Ubuntu 22.04 w/ 6.8.0-40-generic kernel OS for all workloads except LAMMPS, HPCG, NAMD, OpenFOAM, Gromacs which utilize 24.04 w/ 6.8.0-40-generic kernel.
SPEC® and SPECrate® are registered trademarks for Standard Performance Evaluation Corporation. Learn more at spec.org.
7 9xx5-006: AMD internal testing as of 09/01/2024, on FFMPEG (Raw to VP9, 1080P, 302 Frames, 1 instance/thread, video source: https://media.xiph.org/video/derf/y4m/ducks_take_off_1080p50.y4m).
System Configurations: 2P AMD EPYC™ 9965 reference system (2 x 192C) 1.5TB 24x64GB DDR5-6400 running at 6000MT/s, SAMSUNG MZWLO3T8HCLS-00A07, NPS=4, Ubuntu 22.04.3 LTS, Kernel Linux 5.15.0-119-generic, BIOS RVOT1000C (determinism enable=power), 10825484.25 Frames/Hour Median
2P AMD EPYC™ 9654 production system (2 x 96C) 1.5TB 24x64GB DDR5-5600, , SAMSUNG MO003200KYDNC, NPS=4, Ubuntu 22.04.3 LTS, Kernel Linux 5.15.0-119-generic, BIOS 1.56 (determinism enable=power) , 5154133.333 Frames/Hour Median
2P Intel Xeon Platinum 8592+ production system (2 x 64C) 1TB 16x64GB DDR5-5600, 3.2 TB NVME, Ubuntu 22.04.3 LTS, Kernel Linux 6.5.0-35-generic), BIOS ESE122V-3.10, 2712701.754 Frames/Hour Median
For 3.99x the performance with the AMD EPYC 9965 vs Intel Xeon Platinum 8592+ systems
For 1.90x the performance with the AMD EPYC 9654 vs Intel Xeon Platinum 8592+ systems
Results may vary based on factors including but not limited to BIOS and OS settings and versions, software versions and data used.
8 9xx5-022: Source: https://www.amd.com/content/dam/amd/en/documents/epyc-technical-docs/performance-briefs/amd-epyc-9005-pb-gromacs.pdf
9 9xx5-071: VMmark® 4.0.1 host/node FC SAN comparison based on “independently published” results as of 10/10/2024.
Configurations:
2 node, 2P AMD EPYC 9575F (128 total cores) powered server running VMware ESXi8.0 U3, 3.31 @ 4 tiles,
https://www.infobellit.com/BlueBookSeries/VMmark4-FDR-1003
2 node, 2P AMD EPYC 9554 (128 total cores) powered server running VMware ESXi 8.0 U3, 2.64 @ 3 tiles,
https://www.infobellit.com/BlueBookSeries/VMmark4-FDR-1002
2 node, 2P Intel Xeon Platinum 8592+ (128 total cores) powered server running VMware ESXi 8.0 U3, 2.06 @ 2.4 Tiles,
https://www.infobellit.com/BlueBookSeries/VMmark4-FDR-1001
VMmark is a registered trademark of VMware in the US or other countries.
10 9xx5-012: TPCxAI @SF30 Multi-Instance 32C Instance Size throughput results based on AMD internal testing as of 09/05/2024 running multiple VM instances. The aggregate end-to-end AI throughput test is derived from the TPCx-AI benchmark and as such is not comparable to published TPCx-AI results, as the end-to-end AI throughput test results do not comply with the TPCx-AI Specification.
2P AMD EPYC 9965 (384 Total Cores), 12 32C instances, NPS1, 1.5TB 24x64GB DDR5-6400 (at 6000 MT/s), 1DPC, 1.0 Gbps NetXtreme BCM5720 Gigabit Ethernet PCIe, 3.5 TB Samsung MZWLO3T8HCLS-00A07 NVMe®, Ubuntu® 22.04.4 LTS, 6.8.0-40-generic (tuned-adm profile throughput-performance, ulimit -l 198096812, ulimit -n 1024, ulimit -s 8192), BIOS RVOT1000C (SMT=off, Determinism=Power, Turbo Boost=Enabled)
2P AMD EPYC 9755 (256 Total Cores), 8 32C instances, NPS1, 1.5TB 24x64GB DDR5-6400 (at 6000 MT/s), 1DPC, 1.0 Gbps NetXtreme BCM5720 Gigabit Ethernet PCIe, 3.5 TB Samsung MZWLO3T8HCLS-00A07 NVMe®, Ubuntu 22.04.4 LTS, 6.8.0-40-generic (tuned-adm profile throughput-performance, ulimit -l 198096812, ulimit -n 1024, ulimit -s 8192), BIOS RVOT0090F (SMT=off, Determinism=Power, Turbo Boost=Enabled)
2P AMD EPYC 9654 (192 Total cores) 6 32C instances, NPS1, 1.5TB 24x64GB DDR5-4800, 1DPC, 2 x 1.92 TB Samsung MZQL21T9HCJR-00A07 NVMe, Ubuntu 22.04.3 LTS, BIOS 1006C (SMT=off, Determinism=Power)
Versus 2P Xeon Platinum 8592+ (128 Total Cores), 4 32C instances, AMX On, 1TB 16x64GB DDR5-5600, 1DPC, 1.0 Gbps NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84 TB KIOXIA KCMYXRUG3T84 NVMe, , Ubuntu 22.04.4 LTS, 6.5.0-35 generic (tuned-adm profile throughput-performance, ulimit -l 132065548, ulimit -n 1024, ulimit -s 8192), BIOS ESE122V (SMT=off, Determinism=Power, Turbo Boost = Enabled)
Results:
CPU Median Relative Generational
Turin 192C, 12 Inst 6067.531 3.775 2.278
Turin 128C, 8 Inst 4091.85 2.546 1.536
Genoa 96C, 6 Inst 2663.14 1.657 1
EMR 64C, 4 Inst 1607.417 1 NA
Results may vary due to factors including system configurations, software versions and BIOS settings. TPC, TPC Benchmark and TPC-C are trademarks of the Transaction Processing Performance Council.
11 9xx5-009: Llama3.1-8B throughput results based on AMD internal testing as of 09/05/2024.
Llama3-8B configurations: IPEX.LLM 2.4.0, NPS=2, BF16, batch size 4, Use Case Input/Output token configurations: [Summary = 1024/128, Chatbot = 128/128, Translate = 1024/1024, Essay = 128/1024, Caption = 16/16].
2P AMD EPYC 9965 (384 Total Cores), 6 64C instances 1.5TB 24x64GB DDR5-6400 (at 6000 MT/s), 1 DPC, 1.0 Gbps NetXtreme BCM5720 Gigabit Ethernet PCIe, 3.5 TB Samsung MZWLO3T8HCLS-00A07 NVMe®, Ubuntu® 22.04.3 LTS, 6.8.0-40-generic (tuned-adm profile throughput-performance, ulimit -l 198096812, ulimit -n 1024, ulimit -s 8192) , BIOS RVOT1000C, (SMT=off, Determinism=Power, Turbo Boost=Enabled), NPS=2
2P AMD EPYC 9755 (256 Total Cores), 4 64C instances , 1.5TB 24x64GB DDR5-6400 (at 6000 MT/s), 1DPC, 1.0 Gbps NetXtreme BCM5720 Gigabit Ethernet PCIe, 3.5 TB Samsung MZWLO3T8HCLS-00A07 NVMe®, Ubuntu 22.04.3 LTS, 6.8.0-40-generic (tuned-adm profile throughput-performance, ulimit -l 198096812, ulimit -n 1024, ulimit -s 8192), BIOS RVOT1000C (SMT=off, Determinism=Power, Turbo Boost=Enabled), NPS=2
2P AMD EPYC 9654 (192 Total Cores) 4 48C instances , 1.5TB 24x64GB DDR5-4800, 1DPC, 1.0 Gbps NetXtreme BCM5720 Gigabit Ethernet PCIe, 3.5 TB Samsung MZWLO3T8HCLS-00A07 NVMe®, Ubuntu® 22.04.4 LTS, 5.15.85-051585-generic (tuned-adm profile throughput-performance, ulimit -l 1198117616, ulimit -n 500000, ulimit -s 8192), BIOS RVI1008C (SMT=off, Determinism=Power, Turbo Boost=Enabled), NPS=2
Versus 2P Xeon Platinum 8592+ (128 Total Cores), 2 64C instances , AMX On, 1TB 16x64GB DDR5-5600, 1DPC, 1.0 Gbps NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84 TB KIOXIA KCMYXRUG3T84 NVMe®, Ubuntu 22.04.4 LTS 6.5.0-35-generic (tuned-adm profile throughput-performance, ulimit -l 132065548, ulimit -n 1024, ulimit -s 8192), BIOS ESE122V (SMT=off, Determinism=Power, Turbo Boost = Enabled).
Results:
CPU 2P EMR 64c 2P Turin 192c 2P Turin 128c 2P Genoa 96c
Average Aggregate Median Total Throughput 99.474 193.267 182.595 138.978
Competitive 1 1.943 1.836 1.397
Generational NA 1.391 1.314 1
Results may vary due to factors including system configurations, software versions and BIOS settings.
12 9xx5-087: As of 10/10/2024; this scenario contains several 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.
Referencing 9XX5-056A: “2P AMD EPYC 9575F powered server and 8x AMD Instinct MI300X GPUs running Llama3.1-70B select inference workloads at FP8 precision vs 2P Intel Xeon Platinum 8592+ powered server and 8x AMD Instinct MI300X GPUs has ~8% overall throughput increase across select inference use cases” and 8763.52 tokens/s (9575F) versus 8,048.48 tokens/s (8592+) at 128 input / 2048 output tokens, 500 prompts for 1.089x the tokens/s or 715.04 more tokens/s.
1 Node = 2 CPUs and 8 GPUs.
