CEO Jensen Heyang made a series of announcements during his Computex keynote, including details about the company’s upcoming DGX supercomputer. Given the direction the industry is clearly heading in, it should come as no surprise that the DGX GH200 is largely aimed at helping companies develop models.
The supercomputer uses the new NVLink Switch system to enable the 256 GH200 Grace Hopper super chips to act as a single GPU (each chip contains an ARM-based Grace CPU and an H100 Tensor Core GPU). This, according to NVIDIA, allows the DGX GH200 to deliver 1 exaflop of performance and get 144 TB of shared memory. The company says that’s 500 times more memory than what you’d find in a single DGX A100 system.
For comparison, the Top500 supercomputers are listed as the only known exascale, having reached approximately 1.2 exaflops on the Linmark scale. This is more than double the peak performance of Japan’s second-place system.
In fact, NVIDIA claims to have developed a supercomputer that can stand shoulder to shoulder with the most powerful system known on the planet (Meta is building a system that it claims will be the fastest AI supercomputer in the world once it’s fully built). NVIDIA says the DGX GH200 architecture delivers 10x more bandwidth than the previous generation, “delivering the power of an AI supercomputer with the simplicity of programming a single GPU.”
Some big names are interested in the DGX GH200. Google Cloud, Meta, and Microsoft should be among the first companies to gain access to the supercomputer to test how it handles AI workloads. NVIDIA says the DGX GH200 supercomputer will be available by the end of 2023.
The company is also building its own supercomputer, Helios, which combines four DGX GH200 systems. NVIDIA expects Helios to be online by the end of the year.
Huang discussed other generative AI developments during his keynote, including developments on the gaming front. The NVIDIA Avatar Cloud Engine (ACE) for games is a service developers will be able to leverage to create custom speech, conversation, and animation AI models. NVIDIA says its ACE gaming software can “give NPCs conversational skills so they can respond to questions with lifelike personalities that develop.”