NVIDIA GH200 72 Core Grace CPU Performance vs. AMD Ryzen Threadripper Workstations

Written by Michael Larabel in Computers on 20 February 2024 at 10:55 AM EST. Page 4 of 5. 29 Comments.
Timed Godot Game Engine Compilation benchmark with settings of Time To Compile. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.
Timed LLVM Compilation benchmark with settings of Build System: Ninja. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.
Timed Gem5 Compilation benchmark with settings of Time To Compile. System76 Thelio Major r5 - Threadripper 7980X was the fastest.
Timed Node.js Compilation benchmark with settings of Time To Compile. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.

For simple code compilation workloads with the job count matching the CPU thread count, the Threadripper workstations were much faster at compiling various open-source software packages.

Graph500 benchmark with settings of Scale: 26. GPTshop.ai - NVIDIA GH200 was the fastest.
Graph500 benchmark with settings of Scale: 26. GPTshop.ai - NVIDIA GH200 was the fastest.
Graph500 benchmark with settings of Scale: 26. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.
Graph500 benchmark with settings of Scale: 26. GPTshop.ai - NVIDIA GH200 was the fastest.

The GH200 Grace CPU performed exceptionally well for the Graph500 HPC benchmark.

OpenVINO benchmark with settings of Model: Face Detection FP16, Device: CPU. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.
OpenVINO benchmark with settings of Model: Person Detection FP16, Device: CPU. HP Z6 G5 A - Threadripper PRO 7995WX was the fastest.

For workloads like OpenVINO AI toolkit for CPU-based AI performance, they weren't as well optimized on AArch64 as x86_64. The real interesting showdown though will be when incorporating the Hopper GPU into compatible AI benchmarks for a follow-up article when having more time with the GH200 again.

OpenVINO benchmark with settings of Model: Face Detection FP16, Device: CPU. GPTshop.ai - NVIDIA GH200 was the fastest.
OpenVINO benchmark with settings of Model: Vehicle Detection FP16, Device: CPU. GPTshop.ai - NVIDIA GH200 was the fastest.
OpenVINO benchmark with settings of Model: Road Segmentation ADAS FP16, Device: CPU. GPTshop.ai - NVIDIA GH200 was the fastest.

But for some OpenVINO workloads the latencies were much lower on the GH200 thanks to the available memory bandwidth.


Related Articles