Deep Learning Benchmark Studies on an Advanced AI Engineering Testbed from the Open Compass Project
Mei-Yu Wang, Julian A. Uran, Paola A. Buitrago · Practice and Experience in Advanced Research Computing · 2023
We present the Open Compass project’s pilot deep learning benchmark results with various AI accelerators. Those accelerators are NVIDIA V-100 and A-100, AMD MI100, as well as emerging novel accelerators such as Cerebras CS-2 and Graphcore. We evaluate their performance on various deep learning training tasks. We then discuss key insights from our experiments and share experiences about evaluating and integrating those novel AI accelerators with our supercomputing systems.