The original tests for computer vision have been augmented with tests covering recommendation systems, natural language understanding, speech recognition and medical imaging.
The Nvidia A100, introduced earlier this year, increased its lead on the ResNet-50 test, beating CPUs by 30x versus 6x in the previous round.
It also outperformed the latest CPUs by up to 237x in the recommender test.
Nvida said this means one DGX A100 system can provide equivalent performance to approximately 1,000 dual-socket CPU servers at much lower cost.
It's not just the A100: Nvidia T4 GPUs beat CPUs by up to 28x in the same tests, and the Nvidia Jetson AGX Xavier led SoC-based edge devices in terms of performance.
11 Nvidia partners submitted a total of 1,029 results using Nvidia GPUs, making up at least 85% of submissions in the data centre and edge categories.
"We're at a tipping point as every industry seeks better ways to apply AI to offer new services and grow their business," said Nvidia general manager and vice president of accelerated computing Ian Buck.
"The work we've done to achieve these results on MLPerf gives companies a new level of AI performance to improve our everyday lives."