Nvidia made several announcements at the International Supercomputing Conference in Hamburg, Germany, this week, including the unveiling of Venado, a new supercomputer built in partnership with HPE at Los Alamos National Laboratory (LANL) that will be the first to use Nvidia’s Grace Hopper Superchip, which Nvidia announced in March.
Grace Hopper will allow the lab to run workloads up to 3x faster than prior GPUs, and also by employing the Grace CPU, LANL will be able to achieve twice the performance per watt of traditional CPUs on “a long tail of unaccelerated applications,” said Dion Harris, head of datacenter product marketing for HPC, AI, and Magnum IO at Nvidia.
“It's projected to deliver over 10 exaflops of AI performance,” he said. “We’re in the era of exascale computing. We're really excited about it because this will not just be the first Grace Hopper system. We're known for accelerated computing will also have a cluster of Grace CPU superchip modules so [LANL] can really leverage a true heterogeneous environment that will be built on our platform, and which will allow them to use the same programming model across both and get optimum performance across not just their GPU-accelerated apps, but that long tail of non-CPU-accelerated apps.”
Harris added that for LANL Venado will be able to help fuel breakthroughs in numerous research areas such as material science, renewable energy and energy distribution,among others.
Among other ISC announcements, Harris talked about how industrial users are using Nvidia’s Omniverse platform on high-performance computing systems to build digital twins that allow them to research, simulate, experiment with and manage models to help them learn things that might otherwise might cost them a fortune to understand using real-word systems.
For example, the U.K. Atomic Energy Authority and the University of Manchester are using Omnivers to build a digital twin of a fusion reactor. “The reason why that's important is that it is an extremely expensive, extremely large undertaking to build a fusion reactor in terms of size, space and money, so you want to make sure that you can build it right.”
He added, “Omniverse allows them to build a digital twin of the actual structure itself, to see a representation of the physical elements of the physical space of a room as it will be real time… so that they can really understand all the intricate details of the system itself. And they can see how the simulation of the fusion lab itself interacts with the actual environment. They also plan to simulate the physics of the plasma containment itself, which is really one of the Holy Grails of fusion energy and having it be sustainable.”
Further details of other Nvidia announcements from ISC can be found on this blog.