What Is The News About?
The use of NVIDIA Grace HopperTM Superchips by nine new supercomputers around the world is speeding up scientific discovery and research, according to NVIDIA. Powering these systems is artificial intelligence (AI), which is causing a sea change in the HPC industry. The merged systems provide 200 exaflops, or 200 quintillion calculations per second, of artificial intelligence capacity.
Developed on Eviden’s BullSequana XH3000 technology, the EXA1-HE supercomputer was handed over to CEA, the French Commission for Alternative Energies and Atomic Energy, in April. Among the Atos Group’s members is Eviden. With 477 Grace Hopper-based compute nodes, the EXA1-HE is part of the BullSequana XH3000 architecture, which offers a new, proprietary warm-water cooling mechanism. An increasing number of NVIDIA Arm-based supercomputers utilize Grace CPU Superchips and the Grace Hopper platform. Systems at the Texas Advanced Computing Center and Los Alamos National Laboratory in the United States, as well as Isambard-AI and Isambard 3 from the University of Bristol in the United Kingdom, are involved in this trend.
Why Is This News Important?
The upcoming online debut of the following new supercomputers based on the Grace Hopper architecture: EXA1-HE in France, constructed by CEA and Eviden; Helios in Poland, by Academic Computer Centre Cyfronet, supported by Hewlett Packard Enterprise (HPE); Alps at the Swiss National Supercomputing Centre, also supported by HPE; JUPITER at the Jülich Supercomputing Centre, Germany; DeltaAI at the National Center for Supercomputing Applications, UIC; and Miyabi at Japan’s Joint Center for Advanced High-Performance Computing, a partnership between the University of Tsukuba and the University of Tokyo’s Information Technology Center, UIC, and the University of Tokyo’s Center for Computational Sciences, UIT, and the University of Tokyo’s Center for Information Technology Center.
Phase one of Isambard-AI includes the HPE Cray EX2500, one of the most efficient supercomputers ever built. It contains 168 NVIDIA GH200 Superchips. Upon the arrival of the remaining 5,280 NVIDIA Grace Hopper Superchips at the National Composites Centre at the University of Bristol this summer, performance will be enhanced by about 32 times. Their accelerated computing platform is comprised of graphics processing units (GPUs) developed on the NVIDIA Hopper architecture, central processing units (CPUs) and Hopper Superchips, Quantum-2 InfiniBand networking, and an extensive suite of artificial intelligence (AI) and high-performance computing (HPC) software.
Wrapping Up
Investment in data, infrastructure, and workforces hosted and owned domestically is being made by countries around the world to foster innovation. This trend is referred to as “sovereign AI.” This has led to an upsurge in interest in developing cutting-edge AI-powered supercomputers. Global scientific supercomputers rely on the GH200 as their central processing unit. It makes use of NVIDIA NVLink®-C2C connection technology and integrates the NVIDIA HopperTM GPU architectures with the Arm-based NVIDIA Grace CPU. Several research centers aim to implement the system and begin conducting real scientific experiments within months, rather than years.
[To share your insights with us as part of editorial or sponsored content, please write to sghosh@martechseries.com]
Comments are closed.