Nvidia has announced its next-generation AI chip platform that uses the new Grace Hopper Superchip with the world’s first HBM3e processor.
The new NVIDIA GH200 Grace Hopper platform is designed for heavy generative AI workloads such as large language models in favor of ChatGPT. Nvidia also mentions other AI applications it can handle, such as recommendation systems and vector databases.
The dual configuration offers 3.5x more memory and 3x more bandwidth than the current generation. It features a single server with 144 Arm Neoverse cores, eight petaflops of AI performance, and 282 GB of HBM3e memory technology.
The next generation
Jensen Huang, CEO of Nvidia, noted that the upgrades “improve throughput, (add) the ability to connect GPUs to bundle performance without compromise, and (have) a server design that can be easily deployed across the data center. implemented.”
The Grace Hopper Superchip can be connected to other Superchips via Nvidia NVLink, to increase the computing power required by deploying the massive models of generative AI.
With this kind of connection, the GPU has full access to the CPU memory, which offers a combined memory of 1.2 TB in dual configuration mode.
The new HBM3e memory is 50% faster than HBM3, with a total combined bandwidth of 10TB/sec, allowing it to run models 3.5x larger. Performance is also improved thanks to the 3x greater bandwidth.
The new Grace Hopper Superchip platform with HBM3e is fully compatible with the Nvidia MGX server specification, which allows any manufacturer to add Grace Hopper to more than a hundred server variations in a fast and cost-effective manner, the company claims.
The news points to Nvidia’s continued dominance in the AI hardware space, as the A100 GPUs were used to power the machines behind ChatGPT, the infamous chatbot that kickstarted the brave new world of advanced automated computing. It then followed this up with its successor, the H100.
Nvidia also says it expects system makers to make the first models based on Grace Hopper in Q2 of 2024.
- Do you have heavy workloads yourself? These are the best workstations you can get right now