Skip to content

El Capitan’s Little Brother Tuolumne Can Conquer Most Prime 10 Supercomputers

Lawrence Livermore Nationwide Laboratory (LLNL) began to put in its El Capitan supercomputer that guarantees to realize computational efficiency of over 2 FP64 ExaFLOPS for labeled nationwide safety analysis. Parallel to this, the LLNL additionally plans to introduce a much less potent however nonetheless extraordinarily quick supercomputer named Tuolumne. This companion mannequin might be devoted to unclassified analysis and can supply 10% to fifteen% of El Capitan’s computational prowess, which continues to be sufficient to beat a lot of the programs within the present Top500 supercomputer listing.

Whereas LLNL’s El Capitan won’t be the primary exascale system and won’t even be the primary to interrupt 2 FP64 ExaFLOPS efficiency file, it’s nonetheless a really particular system as it is going to be the primary exascale supercomputer to make use of AMD’s Intuition MI300A accelerated processing items comprising of Zen 4 x86 cores and CDNA 3-based compute GPUs. When El Capitan goes on-line, it’s going to considerably outpace Frontier, the world’s quickest PC, which is rated for an Rpeak efficiency of 1.679 FP64 ExaFLOPS.

Tuolumne is predicted to make use of related structure to El Capitan, so it’s going to additionally use Intuition MI300A APU. Even with only a tenth of El Capitan’s efficiency, Tuolumne will be a part of the higher echelons of the worldwide supercomputers listing, able to delivering 200 PetaFLOPS. This potential efficiency may place it among the many Prime 10 supercomputers listed within the present Prime 500 listing. Nonetheless, if it manages to ship 15% of El Capitan’s efficiency, it may rival Leonardo, a supercomputer powered by Intel Xeon Platinum 8358 and Nvidia A100 with an Rpeak efficiency of 304.47 PetaFLOPS, which is at the moment ranked because the fourth quickest supercomputer.

“We’re planning to get an unclassified system that might be referred to as Tuolumne,” mentioned Bronis R. de Supinski in an interview with ExaScaleProject.org. De Supinski is the chief expertise officer for Livermore Computing at LLNL. “Will probably be roughly between 10% to fifteen% the dimensions of El Capitan.”

Supercomputers with labeled standing, like the prevailing Sierra and the deliberate El Capitan at LLNL, primarily serve nationwide safety wants. In line with Bronis R. de Supinski, El Capitan might be primarily used for the stockpile stewardship program, guaranteeing the reliability of nuclear weapons with out resorting to actual testing. Conversely, unclassified supercomputers, just like the upcoming Tuolumne, serve various computational necessities spanning varied scientific fields, from analysis and engineering simulations to knowledge evaluation and climate prediction.

“Tuolumne might be contributing extra to the broader vary of scientific areas,” de Supinski added. “There’s lots of supplies modeling. We’ve got usually had a variety of molecular dynamics. Some QCD get run on the system, seismic modeling. What’s going to in all probability occur is that you realize, these kinds of functions, local weather, and that kind of factor will run on Tuolumne. And if there’s a specific case to be made, we are able to often present for briefer runs on the large system.”

Supply: ExaScaleProject.org