AMD EPYC Milan Still Gives Intel Sapphire Rapids Tough Competition In The Cloud

Written by Michael Larabel in Processors on 30 March 2023 at 10:30 AM EDT. Page 7 of 8. 10 Comments.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 16, Model: ResNet-50. c3-highcpu-8 SPR was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 32, Model: ResNet-50. c3-highcpu-8 SPR was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 64, Model: ResNet-50. c3-highcpu-8 SPR was the fastest.
TensorFlow benchmark with settings of Device: CPU, Batch Size: 64, Model: ResNet-50. c3-highcpu-8 SPR was the fastest.

The Intel VMs tested in Google Cloud easily led over the EPYC Milan VMs with the TensorFlow deep learning software. However, that's not really a surprise and is one of the areas where EPYC Genoa is much more competitive thanks to finally adopting AVX-512.

Neural Magic DeepSparse benchmark with settings of Model: NLP Document Classification, oBERT base uncased on IMDB, Scenario: Asynchronous Multi-Stream. t2d-standard-8 AMD was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90, Scenario: Asynchronous Multi-Stream. t2d-standard-8 AMD was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: NLP Question Answering, BERT base uncased SQuaD 12layer Pruned90, Scenario: Asynchronous Multi-Stream. c3-highcpu-8 SPR was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: CV Classification, ResNet-50 ImageNet, Scenario: Asynchronous Multi-Stream. t2d-standard-8 AMD was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: CV Classification, ResNet-50 ImageNet, Scenario: Asynchronous Multi-Stream. c3-highcpu-8 SPR was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: NLP Text Classification, DistilBERT mnli, Scenario: Asynchronous Multi-Stream. t2d-standard-8 AMD was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: NLP Text Classification, DistilBERT mnli, Scenario: Asynchronous Multi-Stream. c3-highcpu-8 SPR was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: CV Segmentation, 90% Pruned YOLACT Pruned, Scenario: Asynchronous Multi-Stream. t2d-standard-8 AMD was the fastest.
Neural Magic DeepSparse benchmark with settings of Model: CV Segmentation, 90% Pruned YOLACT Pruned, Scenario: Asynchronous Multi-Stream. c3-highcpu-8 SPR was the fastest.

The DeepSparse AI software was performing well on the EPYC Milan VMs even without AVX-512 while Genoa will be much more competitive as I have shown in my bare metal EPYC 9004 series testing. The EPYC Milan throughput with DeepSparse was in good shape but the latency was much higher than the Intel Xeon VMs.

OpenCV benchmark with settings of Test: Core. c2d-highcpu-8 AMD was the fastest.
OpenCV benchmark with settings of Test: Graph API. c2d-highcpu-8 AMD was the fastest.
OpenCV benchmark with settings of Test: Image Processing. t2d-standard-8 AMD was the fastest.
OpenCV benchmark with settings of Test: Object Detection. t2d-standard-8 AMD was the fastest.

The AMD EPYC Milan VMs were leading when it came to the OpenCV computer vision benchmarks.


Related Articles
Ampere Computing 2024 Roadmap Update: 256 Core 3nm CPU In 2025
Benchmarking The First RISC-V Cloud Server: Scaleway EM-RV1 Performance
AMD Announces Ryzen PRO 8840 & PRO 8000G Series CPUs
Intel Announces Gaudi 3 AI Accelerator, Intel Xeon 6 Brand
AMD Zen 4 vs. Zen 4C Performance, Zen 4C Core Scaling With Ryzen 5 8500G
Intel Core i3 14100 / i5 14500 vs. AMD Ryzen 5 8500G / 8600G In 500+ Benchmarks