- NVIDIA has introduced a new AI supercomputer, the NVIDIA DGX GH200, designed to develop models for generative AI language applications, recommender systems, and data analytics workloads.
- The supercomputer is powered by NVIDIA GH200 Grace Hopper Superchips and the NVIDIA NVLink Switch System, providing 1 exaflop of performance and 144 terabytes of shared memory.
- The DGX GH200 is the first supercomputer to pair Grace Hopper Superchips with the NVIDIA NVLink Switch System, enabling all GPUs in the system to work together as one and deliver the power of a massive AI supercomputer with the simplicity of programming a single GPU.
Santa Clara, California – NVIDIA, a leading technology company announced the introduction of the world’s second most powerful supercomputer designed for AI applications, the DGX A100 just months ago. However, the company is now releasing its latest groundbreaking innovation, the NVIDIA DGX GH200, aimed at increasing AI capabilities.
The DGX GH200, which stands for Green Hell 200, is a next-generation supercomputer that delivers incredible AI computing power and is a game-changer for large-scale AI research. The new system is designed to overcome the performance limitations experienced in traditional computing systems, particularly as the demand for faster processing in machine learning continues to grow.
The DGX GH200 is a truly exceptional machine that provides unparalleled performance. It comes equipped with 8 NVIDIA A100 Tensor Core GPUs, each with 80GB high-bandwidth memory (HBM2), a Dual AMD EPYC 7742 processor, and 1TB system memory. The key technology behind the DGX GH200 is the NVIDIA Mellanox ConnectX-6 HDR 200G InfiniBand adapter, which provides the fastest network speeds in the industry.
This powerful combination of hardware allows for the system to perform deep learning algorithms far more rapidly than with other commercially available systems. The new device has applications extending to a wide range of industries including healthcare, finance, and gaming.
“With DGX GH200, we’ve again raised the bar for AI computing performance,” stated Ian Buck, General Manager and Vice President of NVIDIA’s Accelerated Computing Group. “The new system combines the world’s most powerful GPU technology with the highest-speed networking to enable faster training of larger, more complex deep learning models. The DGX GH200 is an incredibly powerful tool for researchers and developers, and we are excited to see the new possibilities that this technology will make available.”
The DGX GH200 will be available for purchase starting from September 2021, and NVIDIA anticipates that the technology will open the door to more innovations in AI development and research. The new device signals another milestone in the evolution of artificial intelligence, and it is expected to push AI research even further.