In the rarefied (and carefully cooled) air of the world’s top supercomputers, a new leader has emerged. According to the T0P500 list, the El Capitan, running AMD Epyc CPUs and Instinct accelerators, took the top spot in its supercomputer rankings. It shouldered Frontier, which also uses AMD Epyc CPUs, to second place. Third place now belongs to Aurora, which uses Intel Xeon CPU Max processors and Intel Data Center GPU Max accelerators.
Lawrence Livermore National Laboratory (LLNL) in Livermore, CA is home to El Capitan. The supercomputer took the top spot by earning a High-Performance Linpack (HPL) score of 1.742 exaflops. El Capitan has Epyc CPUs and Instinct MI300A accelerators. As the Big Ideas Lab podcast noted, El Capitan can perform more than 2 quintillion calculations per second.
The system is also energy efficient, scoring 18th place on the GREEN500 list. That’s impressive efficiency for the best-performing supercomputer around.
AMD noted that El Capitan is also the first exascale-class system in the National Nuclear Security Administration (NNSA), which provides compute power for the LLNL, Lost Alamos, and Sandia National Laboratories. Among the many projects and duties of the NNSA is one for which we definitely want them to have all the computing power they need: certifying and monitoring our country’s aging nukes.
“Prior to 1992, we would go off to Nevada, drill a big hole in the ground, put the weapon down there and set it off,” said Rob Neely, LLNL’s Associate Director for Weapon Simulation and Computing to Big Ideas Lab. Once those tests ended, “That really spearheaded a big push in the United States to use supercomputing as one leg of a new tool called Science-Based Stockpile Stewardship, designed to make sure we could retain our confidence in these weapons.”
Both El Capitan and (now 2nd place) Frontier supercomputers use the Cray Slingshot 11 network for data transfer. Frontier is located in the Oak Ridge National Laboratory in Tennessee and moved up in performance in the new T0P500 list—just not enough to keep the number one spot. Previously, Frontier’s HPL score was 1.206 Elop/s; it’s now 1.353 Eflop/s. That’s due to a massive increase in its total cores, which now stands at 9,066,176.
Moving from second place to third is Aurora, which is based in the Argonne Leadership Computing Facility in Illinois. The Intel Xeon-based supercomputer delivered 1.012 Exaflop/s.
#World039s #Fastest #Supercomputer #AMD #Instinct #Chips
source: https://www.extremetech.com/computing/the-worlds-fastest-supercomputer-has-amd-instinct-chips
