MLPerf Inference v2.1 Results Announced For Datacenter & Edge Scale Benchmarks
The results for MLPerf Inference v2.1, including datacenter and edge scales, are in, seeing greater participation than previous rounds.
Created on September 8|Last edited on September 8
Comment
MLPerf, an AI hardware benchmarking round hosted by MLCommons, puts industry-leading hardware to the test in a variety of machine learning tasks. The benchmarking rounds happen periodically, and today we're getting a look at the latest results on inference tasks for both datacenter and edge scales with MLPerf Inference v2.1.
Like all MLPerfs benchmarking rounds, the tests are standardized and submission is open to all AI hardware system companies who would like to participate. For inference tests, pre-trained models are used and various benchmarks are measured on inference tasks. Here's what each inference setup was benchmarked on for both datacenter and edge scales (excluding the "Commerce" section for edge):
Overall, compared to the previous benchmarking for datacenter- and edge-scale inference tests, participation was up 1.37× for performance results and 1.09× for power measures. Participants for this round include big names including NVIDIA, Azure, Intel, and many others.
Find out more
If you'd like to take a look at the results and see more details for each test scale, you can find them here for datacenter and here for edge.
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.