NVIDIA and the National Energy Research Scientific Computing Center (NESC) have operated the Perlmutter. They described Perlmutter as the world's fastest supercomputer for AI workloads. The supercomputer has 6144 NVIDIA A100 Tensor Core GPUs. He is tasked with putting together the largest 3D map of the visible universe, among other projects.
Nvidia said Perlmutter is the fastest system on the planet at processing workloads using the 16-bit and 32-bit mixed-precision math used in AI applications. Later this year, a second phase will add more AI supercomputing power to Perlmutter. In one project, NVIDIA wrote, a supercomputer is helping to assemble the largest 3D map of the visible universe to date.
The computer processes data from the Dark Energy Spectroscopy (DESI) instrument. It is a type of cosmic camera that can capture up to 5,000 galaxies in a single exposure. For example, researchers need the speed of Perlmutter's GPUs to take dozens of exposures from one night to know where to point DESI the next night. Preparing year-long data for publication could take weeks or months across previous systems. But the Perlmutter should help them get the job done in less than a few days.
Nvidia, artificial intelligence, and machine learning
People are exploring larger and larger models of neural networks, and there is a demand for access to more powerful resources. Perlmutter meets this need for AI with A100 GPUs, a full-flash file system, and data streaming capabilities. Perlmutter gives nearly 7,000 NERSC-supported researchers access to 4 exaFLOPS performance mixed-precision computing for AI-assisted science projects.
In addition to the DESI mapping project, researchers spend time with the supercomputer working in areas such as climatology. Perlmutter helps investigate subatomic reactions to discover green energy sources.
NVIDIA said that this project requires a special blend of artificial intelligence and high-performance computing that Perlmutter provides.
Conventional supercomputers can barely handle the math required to generate simulations of a few atoms over a period of a few nanoseconds using software such as Quantum Espresso. Scientists can study more atoms over longer timescales by combining high-resolution simulations with machine learning. The ability to leverage AI in supercomputing makes researchers optimistic about DESI. The project also aims to shed light on dark energy, the mysterious physics behind the accelerating expansion of the universe.