AMD Investor Day Presentation Deck
Endnotes (Cont.)
MI200-63- Testing Conducted by AMD performance lab as of 25/5/2022 using SuperBench v 0.4.0 benchmark, Resnet 50/101/152 datasets. All configurations used Framework: PyTorch 1.9 EPYC/Instinct system: Dual socket, 64 core, 2nd
Gen AMD EPYC™ 7002 Series CPU powered server with 8x AMD Instinct™ MI250X OAM (128 GB HBM2e) 500W GPUs with AMD Infinity Fabric™ technology, ROCm™ 5.1.0 Benchmark: Resnet model (Median scores of Resnet50, Resnet 101,
Resnet152 datasets) with AMD | Microsoft optimized batch sizes tuned for Resnet results for system configurations that are not yet available upstream. Commit(Container): computecqe/superbench:rocm5.1.3_superbench04 from here
Benchmark Results: Resnet testing resulted in median throughput scores of 8x MI250X: resnet50 = 9708.873, resnet101 = 6705.041, resnet152 = 5250.635 Samples (Throughput)/ sec. Training model separates copies of model on each
GPU; total system throughput obtained by calculating the sum of the throughput obtained on each GPU. Vs. EPYC/Nvidia system: NVIDIA DGXA100, Dual socket 64-core AMD EPYC 7742 Series CPUs, with 8x NVIDIA A100 SXM 80GB
(400W), CUDA 11.6 and Driver Version 510.47.03, OS: Benchmark: Resnet model (Median scores of Resnet50, Resnet 101, Resnet152 datasets) Commit(Container): superbench/superbench:v0.4.0-cuda11.1.1 from here Benchmark Results:
Resnet testing resulted in median throughput of 8x A100: resnet50 = 7057.052, resnet101 = 4968.324, resnet152 = 3806.318 Samples (Throughput)/ sec. Details on SuperBench found here Server manufacturers may vary configurations,
yielding different results. Performance may vary based on use of latest drivers and optimizations MI200-63
MI300-03 - Measurements by AMD Performance Labs June 4, 2022. MI250X FP16 (306.4 estimated delivered TFLOPS based on 80% of peak theoretical floating-point performance). M1300 FP8 performance based on preliminary
estimates and expectations. Actual results based on production silicon may vary. M1300-03View entire presentation