AMD BIOS Tuning Guide Impact For Boosting AI/ML Performance On EPYC 9005 Series

Written by Michael Larabel in Software on 29 November 2024 at 10:36 AM EST. Page 4 of 4. 6 Comments.

TensorFlow benchmark with settings of Device: CPU, Batch Size: 256, Model: ResNet-50. AI/ML Tuning Recommendations was the fastest.

TensorFlow benchmark with settings of Device: CPU, Batch Size: 512, Model: ResNet-50. AI/ML Tuning Recommendations was the fastest.

LiteRT benchmark with settings of Model: Mobilenet Float. AI/ML Tuning Recommendations was the fastest.

LiteRT benchmark with settings of Model: NASNet Mobile. AI/ML Tuning Recommendations was the fastest.

LiteRT benchmark with settings of Model: SqueezeNet. AI/ML Tuning Recommendations was the fastest.

PyTorch benchmark with settings of Device: CPU, Batch Size: 256, Model: ResNet-50. AI/ML Tuning Recommendations was the fastest.

PyTorch benchmark with settings of Device: CPU, Batch Size: 256, Model: ResNet-152. AI/ML Tuning Recommendations was the fastest.

PyTorch benchmark with settings of Device: CPU, Batch Size: 512, Model: ResNet-50. AI/ML Tuning Recommendations was the fastest.

PyTorch benchmark with settings of Device: CPU, Batch Size: 512, Model: ResNet-152. AI/ML Tuning Recommendations was the fastest.

oneDNN benchmark with settings of Harness: IP Shapes 1D, Engine: CPU. AI/ML Tuning Recommendations was the fastest.

oneDNN benchmark with settings of Harness: Recurrent Neural Network Training, Engine: CPU. AI/ML Tuning Recommendations was the fastest.

oneDNN benchmark with settings of Harness: Recurrent Neural Network Inference, Engine: CPU. AI/ML Tuning Recommendations was the fastest.

ONNX Runtime benchmark with settings of Model: ResNet50 v1-12-int8, Device: CPU, Executor: Standard. AI/ML Tuning Recommendations was the fastest.

XNNPACK benchmark with settings of Model: FP16MobileNetV3Small. AI/ML Tuning Recommendations was the fastest.

XNNPACK benchmark with settings of Model: QS8MobileNetV2. AI/ML Tuning Recommendations was the fastest.

Across a range of AI / machine learning software tested, following the AMD EPYC 9005 BIOS and Workload Tuning Guide proved worthwhile for further enhancing the performance of the new EPYC Turin processors. The BIOS defaults are already in great shape for delivering excellent generational EPYC uplift and battling the Intel Xeon / Arm server competition, but for those deploying 5th Gen EPYC servers for specific use-cases/workloads, this tuning guide is helpful for better optimizing the platform for the greatest potential. Following the AMD BIOS and workload tuning guide is quite easy and quick while depending upon the workload can yield some nice benefits.

I will be working through additional workload tuning guide recommendations around Java, relational database servers, HPC, and more, but those wanting to go through the guide right now can find it on AMD.com.

6 Comments

If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.

Page: 1 2 3 4