Vera Rubin delivers up to 2× GB300’s training performance per GPU in NVIDIA’s MLPerf Training 6.1 results; its DeepSeek-V3 run scaled from 256 to 1,024 GPUs at about 97% efficiency. These are benchmark results, not an across-the-board speedup.

Published
NVIDIA / MLPerf Training 6.1 results ↗ · Source dated
—












Checking account…
Loading comments…