Nvidia Corp.’s latest artificial intelligence platforms have once again set new performance records in the latest MLPerf benchmarks.

MLPerf is an industry benchmarking group that was set up in May 2018 and is backed by companies that include Amazon.com Inc., Baidu Inc., Facebook Inc., Google LLC, Intel Corp. and Microsoft Corp., as well as Harvard and Stanford universities. The tests are meant to establish how fast different AI systems can perform inference, or come to a conclusion or result based on the information they digest.

Nvidia said its systems won every test across all six application areas for data center and edge computing systems in the second version of MLPerf Inference. In the first edition of MLPerf there were just two inference tests, for computer vision, but this edition saw new tests added for recommendation systems, natural language understanding, speech recognition and medical imaging.

Nvidia’s AI systems are powered by its latest graphics processing unit, the Nvidia A100 (pictured) that was introduced earlier this year featuring third-generation Tensor Cores and multi-instance GPU technology. The A100 GPU increased its lead on the ResNet-50 image classification test, beating the most advanced central processing units by 30 times, compared with just six times in the last round of tests. In addition, the A100 outperformed CPUs by 237 times in the new recommender test for data center inference.

The company said the results show that a single Nvidia DGX A100 server can achieve the same performance on AI inference as about 1,000 dual-socket CPU servers. That adds up to extreme cost efficiency benefits for companies that use Nvidia’s system to bring their AI models online, the company said.

Nvidia also excelled at the edge, where lightweight devices need a much more energy-friendly and cost-effective platform. The Nvidia T4 Tensor Core GPU is designed just for that purpose, and beat CPUs by up to 28 times in the same series of AI inference tests. Nvidia’s older Jetson AGX Xavier GPU, meanwhile, was still the top performer among system-on-chip-based edge devices.

“We’ve continued to increase our performance leadership,” Paresh Kharya, Nvidia’s senior director of product management and marketing, said in a press briefing.

In a blog post, Kharya said further evidence of the company’s prowess in AI can be seen in the field with customers already running its A100 GPUs and other hardware in production. One area where AI is really making an impact is in medical imaging, he said. For example, the startup Caption Health has deployed Nvidia’s systems to help doctors diagnose the results of echocardiograms and save lives during the early days of the COVID-19 pandemic.

Nvidia’s AI systems are making their presence felt in other industries too, ranging from automotive, cloud, robotics, healthcare, retail and financial services, with companies that include American Express Co., BMW AG, Capital One Financial Corp., Domino’s Pizza, Inc., Ford Motor Co., GE Healthcare, Samsung Electronics Co. Ltd. and Toyota Motor Corp.

Nvidia’s AI platforms have also been integrated with some of the world’s most powerful supercomputers for advanced scientific research purposes, including the University of Florida’s HiPerGator, and four new systems that were just unveiled last week in Italy, Czech Republic, Luxembourg and Slovenia.

With reporting from Robert Hof

Images: Nvidia