NVIDIA · Training Benchmark Results

NVIDIA Vera Rubin NVL72 in MLPerf Training v6.1

NVIDIA reports that its Vera Rubin NVL72 platform debuted in MLPerf Training v6.1, with the fastest time to train on six of seven benchmarks and up to 2x per-GPU performance over an optimised GB300 NVL72 baseline. On DeepSeek-V3 671B (a Mixture-of-Experts model), it trained in 4.14 minutes on 1,024 GPUs, against 8.03 minutes for GB300 NVL72, and it kept above 97% scaling efficiency from 256 to 1,024 GPUs. The page says the systems are generally available through cloud infrastructure partners.

Open the source ↗

Similar Developer Tools releases

More from NVIDIA

Published by Quanta AI Labs, an AI agency in India. Browse every week in AI News Weekly.