Nvidia’s powerful H100 GPU will arrive in October

Increase / A press handout showing the Nvidia H100 Hopper GPU and its applications.


Today GTC conference In a keynote, Nvidia announced that its H100 Tensor Core GPU is in full production and that technical partners such as Dell, Lenovo, Cisco, Atos, Fujitsu, GIGABYTE, Hewlett-Packard Enterprise and Supermicro will begin shipping products based on the H100 , next month.

The H100, part of the “Hopper” architecture, is Nvidia’s most powerful AI-focused GPU ever, surpassing its previous high-end chip, A100. The H100 includes 80 billion transistors and a special “Transformer Engine” to accelerate machine learning tasks. It also supports Nvidia NVLink, which links GPUs for increased performance.

According to Nvidia’s press release, the H100 also delivers efficiency benefits, offering the same performance as the A100 with 3.5 times better energy efficiency, 3 times lower cost of ownership and using 5 times fewer server nodes.

Nvidia expects the H100 chip to be used in a variety of industrial, medical, supercomputing and cloud applications, ranging from large language models, drug discovery, recommender systems, conversational artificial intelligence and more. Based on the track record of the previous GPU with the A100 “Ampere” architecture, analysts believe that the H100 chip is likely to make a big impact in the field of artificial intelligence. It is also very likely to play a role in the next generation image synthesis models.

Nvidia has announced that more than 50 server models based on the H100 from various companies will be on the market by the end of the year. And Nvidia itself will begin integrating the H100 into its own Nvidia DGX H100 enterprise systems that contain eight H100 chips and provide performance of 32 petaflops.

The news that the H100 is in full production and shipping soon can be seen as a relief for Nvidia investors. The firm’s shares fell in early September after it announced that the US government was imposing restrictions to export the H100 to China, where the chip was also partially designed and manufactured. The US has allowed development of the H100 to resume, but export restrictions to China and Russia are still in place.

Source link