Nvidia has unveiled a new DGX GH200 AI supercomputer, underpinned by its new Grace Hopper superchip and qualified towards establishing and supporting large language designs.

“DGX GH200 AI supercomputers combine Nvidia’s most superior accelerated computing and networking technologies to broaden the frontier of AI,” Nvidia CEO Jensen Huang claimed in a website post.

The supercomputer, in accordance to Huang, brings together the company’s GH200 Grace Hopper superchip and Nvidia’s NVLink and Switch Technique, to make it possible for the enhancement of large language products for generative AI language apps, recommender systems, and facts analytics workloads.

Nvidia’s DGX GH200 works by using the NVLink interconnect technology to combine 256 Grace Hopper superchips into a solitary graphics processing device (GPU) in order to extract “1 exaflop of performance and 144 terabytes of shared memory — just about 500 occasions a lot more memory than the prior generation NVIDIA DGX A100, which was introduced in 2020.”

The chip maker will emulate the system it took with its DGX Pods in building the new supercomputer obtainable.

Earlier in March, Huang reported the firm struck a deal to make its DGX units out there by various cloud vendors, somewhat than installing the needed components on-premises.

Presently, Microsoft, Meta, and Google Cloud have obtain to the new supercomputer, the company claimed.

Nvidia also mentioned the new Grace Hopper superchip that fuels the DGX GH200 AI supercomputer is in comprehensive output method and techniques with the superchip are anticipated to be out there afterwards this calendar year.

The firm also claimed that it was making use of the new Grace Hopper superchip to help SoftBank style up coming-generation dispersed info centers that will be capable of managing generative AI and 6G purposes.

These details facilities will be distributed across Japan, the organizations explained in a weblog put up.

Earlier in March, the enterprise had released new knowledge processing models (DPUs) and GPUs, like the BlueField 3 DPU.