Nvidia Announces H200 GPU, Teases Next-gen B100
Nvidia last week announced its HGX H200 GPU, a follow-up of last year’s H100 GPU based on the Nvidia Hopper architecture.
H200 is the first GPU to offer HBM3e memory, which is high capacity and high bandwidth memory, to fuel the acceleration of generative AI and large language models.
According to Nvidia, the H200 delivers 141GB of memory at 4.8 terabytes per second, nearly double the capacity and 2.4 times more bandwidth compared with the Nvidia A100, its predecessor.
Shortages of GPUs have been blamed for hindering the development of new AI models. Beyond manufacturing more GPUs, the other option is to have more powerful ones.
The H200 will be available in Nvidia HGX H200 server boards with four- and eight-way configurations and is both hardware and software compatible with HGX H100 systems.
Nvidia claims the H200 can be deployed in every type of data center, including on-premises, cloud, hybrid-cloud, and edge. Its current global ecosystem of partner server makers can update their existing systems with an H200.
“To create intelligence with generative AI and HPC applications, vast amounts of data must be efficiently processed at high speed using large, fast GPU memory,” said Ian Buck, vice president of hyperscale and HPC at Nvidia.
“With NVIDIA H200, the industry’s leading end-to-end AI supercomputing platform just got faster to solve some of the world’s most important challenges.”
One GPU per year
Separately, TechRadar reported that the next Nvidia GPU, the B100 Blackwell GPU, will more than double the performance of the H200. This was cited based on a slide Nvidia showed at SC23 (The International Conference for High Performance Computing, Networking, Storage, and Analysis).
“The H100 performs 11 times better than the A100, and the H200 performs 18 times better than this chip, with next year’s B100 set to ramp the performance up even higher, according to a chart comparing performance against the 175-billion-parameter GPT-3 large language model (LLM),” said the report.
While a release date of 2024 was cited for the B100, it is more likely an announcement, with a release in 2025 as part of Nvidia’s annual GPU updates.
The H200 will be available from global system manufacturers and cloud service providers starting in the second quarter of 2024.
Paul Mah is the editor of DSAITrends. A former system administrator, programmer, and IT lecturer, he enjoys writing both code and prose. You can reach him at [email protected].
Image credit: Nvidia
