Named after astrophysicist Saul Perlmutter, this supercomputer uses over 6,000 Nvidia A100 Tensor core GPUs to build the most ambitious 3D map of the universe ever seen. The first task is to do.
By Mark jaction
Nvidia has launched a new supercomputer at the National Energy Research Scientific Computer Center (NERSC) in California, which is said to be the fastest AI workload in the world.
Named after astrophysicist Saul Perlmutter, this supercomputer uses over 6,000 Nvidia A100 Tensor core GPUs to build the most ambitious 3D map of the universe ever seen. The first task is to do.
In its current state, Perlmutter can provide nearly four exaflops of AI performance, which Nvidia says is “the fastest system on earth for 16-bit and 32-bit mixed-precision arithmetic AI.” Its performance is further enhanced as part of “Phase 2” with the introduction of a second tranche of CPU cores.
Once the system is complete, it is expected to be among the top five supercomputers in the world. Top 500 A ranking that uses the High Performance Linpack (HPL) benchmark to assess overall performance.
Nvidia supercomputer
According to Nvidia, more than 7,000 researchers plan to bring the Perlmutter system to practical use, with dozens of applications already in place aimed at advancing astrophysics and climate science.
“In one project, a supercomputer helps build the largest 3D map of the universe ever visible. It’s a kind of space camera that can capture as many as 5,000 galaxies with a single exposure, dark energy. Processes data from a spectrometer (DESI). Blog post From Dion Harris, Nvidia HPC & AI Product Marketing Lead.
“Researchers need the speed of Perlmutter’s GPU to capture dozens of exposures overnight to know where to direct DESI the next night. Previous systems would take a year’s worth of data. It took weeks or months to get ready, but with Perlmutter you can complete this task in just a few days. “
Once the map of the universe is complete, it is hoped that it will help researchers learn more about the dark energy behind the accelerated expansion of the universe. The discovery of dark energy in 2011 earned Sole Palmter a Nobel Prize.
Other projects that are configured to run on Perlmutter machines have similar ambitious goals. Many of them take advantage of the unique quality of the A100’s Tensor core to simulate interatomic interactions that were previously nearly impossible.
This research could be enlightening in the field of materials science as it could reveal ways to develop more efficient batteries, biofuels, etc.
Nvidia is creating a 3D map of the universe with the world’s strongest AI supercomputer
Originally published at Texas news today