MLPerf
The industry-standard benchmark suite for AI training and inference performance; the closest thing to apples-to-apples numbers across accelerator vendors.
Current numbers
~1.08 TB/speak single-submission throughput at 92.2% GPU utilization and 9.9M IO/s (Volumez on AWS) in MLPerf Storage v1.0
≥90%accelerator-utilization threshold for a passing MLPerf Storage result (70% for Cosmoflow); holding this line is the loader's core objective
200+ / 26performance results from 26 organizations in MLPerf Storage v2.0; systems now feed ~2x the accelerators of v1.0