The National Energy Research Scientific Computing Center is set to complete the construction of its largest, fastest supercomputer at Lawrence Berkeley National Laboratory, or Berkeley Lab, by spring 2021.
The Perlmutter supercomputer, named after Nobel Prize-winning astrophysicist and UC Berkeley physics professor Saul Perlmutter, is designed to support data analysis, machine learning and large-scale simulations.
“The supercomputer isn’t only helping the scientists at Berkeley, but scientists all over the country,” Perlmutter said. “They’re developing and designing the supercomputer in a way that’s much more effective for the whole country.”
The construction of a new and more powerful supercomputer represents the Department of Energy’s Office of Science’s “commitment to extreme-scale science, developing new energy sources, improving energy efficiency and discovering new materials,” related to research conducted through the DOE, according to Berkeley Lab news.
Through the use of Berkeley Lab’s other supercomputers, Perlmutter’s team has been working to measure the speed of the expansion of the universe through simulations and data analysis.
Kathy Yelick, associate laboratory director for computing sciences at Berkeley Lab, said about 7,000 people nationwide use Berkeley Lab’s supercomputers for research related to subjects such as physics, chemistry, material sciences, biology and the study of climate change.
According to Yelick, the Perlmutter supercomputer has graphics processing units, or GPUs, that are well-suited for machine learning problems and streaming data for experiments straight into the system for the facilitation of faster data analysis. It will also be integrated with cooling infrastructure that will release heat buildup from the immense amount of power surging through the supercomputer system.
“It’s really because either the computational problems are too large for just running on an individual computer, or because the data sets are too large. Those are the two main reasons you use the supercomputer,” Yelick said. “A supercomputer is a little different than a set of other computers, although it’s built out of the same kinds of processors and GPUs. It has a high-speed network so that all the processors can work together to solve a single problem.”
The supercomputer was originally set to be installed in 2020 but was delayed another year because the necessary technology was not yet available, according to Yelick.
Perlmutter said this supercomputer will be better able to tune its own capabilities to the needs of scientists.
“At this stage, there are so many projects in science that depend on being able to do these remarkably fast calculations,” Perlmutter said. “It’s a very nice time for the generation that’s just coming out that’s actually somewhat more dedicated to the scientific data analyses that we all face.”