This site may earn chapter commissions from the links on this page. Terms of use.

Supercomputing wins are a big deal for semiconductor manufacturers. Winning space in top systems is seen as a mark of stability and longevity. It implies vendors at the meridian of the market and with access to important, long-term authorities contracts believe your hardware is robust and capable enough to be used for cut-edge scientific research at one of the United states' largest installations. The economic science of these deals are decidedly more than cloudy, but as a matter of public relations, companies tend to adore them. Both Nvidia and AMD have something to crow about today, given a new announcement from Cray and Berkeley National Labs.

The National Energy Research Scientific Calculating Centre volition use a Cray Shasta installation for its side by side-generation supercomputer, codenamed "Perlmutter." To be clear, "Shasta" is what Cray calls the supercomputer compages, while Perlmutter is the name of the specific NERSC arrangement to be constructed. Shasta is designed to be compatible with both ARM and x86 architectures from Intel and AMD with support for a variety of interconnect standards, including Cray's Slingshot, Intel's Omnipath, and Mellanox (Infiniband). For those of yous who acquaintance Infiniband with Intel, the company bought its ain Infiniband technology back in 2022, just Omnipath is a dissimilar, Intel-specific applied science, which Mellanox'due south Infiniband competes with. Shasta is designed to integrate components and accelerators from a variety of companies, including GPUs, FPGAs, and AI-specific accelerators.

Cray'due south new Slingshot interconnect is a critical part of the system (and scaling interconnect ability downwardly is of import to how we eventually accomplish exascale compute capability). A short video near the new interconnect is embedded below:

Nosotros don't accept figures notwithstanding on which Epyc CPUs or how many cores NERSC will deploy — or how many GPUs it will contain. Nvidia, nonetheless, is likewise crowing nigh its own inclusion on the project, with one study challenge that 50 pct of the workloads Perlmutter volition perform are capable of running on GPUs. This volition be the first NERSC supercomputer focused on heterogeneous compute, so picking up the win is a nice feather in Nvidia's cap and doubtless represents a great deal of long-term endeavour to ensure workloads will run well. Nosotros don't know which Tesla products NERSC will utilise, either, but would expect it to be hardware most the top of Nvidia's Tesla stack.

NERSC-gpu-readiness

NERSC GPU kernels

NERSC does work in a number of disquisitional fields, including nuclear fusion research, climate modeling, materials science research, and biology research focused on molecular construction and how information technology relates to drug discovery and vaccine development. While this machine isn't itself an exascale-class deployment, it's expected to use some of the same technologies and standards we'll deploy for exascale computing when the first systems come online (theoretically) in 2022.

Now read: Japan Tests Silicon for Exascale Computing in 2022, 750 Raspberry Pis Turned Into Supercomputer for Los Alamos National Laboratory, and Fujitsu Enters Deep Learning, AI Markets With Custom Architecture