NVIDIA @ SC22: Widespread H100 & Quantum-2 Adoption; Omniverse & Library Updates
NVIDIA gave us several announcements today, including the widespread adoption of their newest hardware with partnered companies, and improvements to Omniverse and other software solutions, for HPC-scale scientific and industrial AI applications.
Created on November 14|Last edited on November 15
Comment
Today at Super Computing 2022, NVIDIA published a special address outlining a number of their systems from the newest tech for high-performance computer, to software for quantum computing simulation, applications for scientific research, and more.
With this publication came a number of announcements including the widespread adoption of the H100 Tensor Core GPUs and Quantum-2 InfiniBand networking systems, major updates to libraries geared at scientific research, and an expansion to the Omniverse platform also in the effort to support scientific research.
H100 & Quantum-2 adoption and HPC library updates
Many announcements are coming out of SC22, and one of the common threads is that of NVIDIA's hardware being used in a variety of applications. Microsoft Azure came out as the first to put Quantum-2 InfiniBand to use in the public cloud, while ASUS & Dell announced new servers running H100s, Hewlett Packard Enterprise put H100s in their new supercomputers, and many other companies are similarly using the latest NVIDIA hardware for their solutions and research.
To go along with the widespread adoption of NVIDIA's hardware, they've released major updates to libraries present in their HPC package, mostly focusing on utility in scientific research. This includes new features for quantum computing simulation in CUDA and cuQuantum, improve data computing features with DOCA, and more.
Omniverse
The other big announcements had to do with NVIDIA Omniverse, their computing platform for a variety of applications including digital twins, large-scale simulation, and AI applications therein.
Omniverse can now connect directly with scientific data visualization software, and now supports batch rendering on A100 and H100 systems.
They also put a focus on Omniverse's use in HPC systems, such as the NVIDIA OVX server for large-scale and real-time digital twins for science and industry, as well as Omniverse Cloud as an off-site option.
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.