The US supercomputer built with HPE and AMD technology has shot past the competition, surpassing the threshold of a quintillion calculations per second.
The biannual Top 500 supercomputer ranking has named a US machine called Frontier the fastest supercomputer to ever exist and the first officially recognised exascale supercomputer.
In a HPL benchmark test, the Frontier supercomputer at the US Department of Energy’s Oak Ridge National Laboratory (ORNL) gave a performance of 1.1 exaflops. This makes it the first to break the exascale barrier in this test, surpassing the threshold of a quintillion calculations per second.
The machine also hit the top spot on the Green 500 list, which rates the energy use and efficiency of supercomputers, with a performance of 62.68 gigaflops per watt.
Frontier is based on the latest HPE Cray EX architecture and equipped with AMD processors, packing 8.7m cores.
It mixed-precision computing performance was 6.88 exaflops, or more than 6.8 quintillion flops per second. This test measures speeds in computing formats typically used by the machine learning methods that drive advances in artificial intelligence.
To explain the power of an exascale supercomputer, ORNL said that if each person on Earth completed one calculation every second, it would still take more than four years to do what an exascale computer can do in one second.
“Frontier is ushering in a new era of exascale computing to solve the world’s biggest scientific challenges,” ORNL director Thomas Zacharia said. “This milestone offers just a preview of Frontier’s unmatched capability as a tool for scientific discovery.”
Frontier overtook the previous leader Fugaku, which earned the top spot in 2020. Based in the city of Kobe in Japan and co-developed by Riken and Fujitsu, its 7.6m cores allowed it to achieve 442 petaflops in the HPL benchmark. Fugaku’s performance previously put it three times ahead of the next highest spot on the Top 500 list.
ORNL said the next steps for Frontier include continued testing and validation of the system. The machine is expected to achieve final acceptance and early science access later this year, and full science access at the beginning of 2023.
Speaking to SiliconRepublic.com this week, Prof JC Desplat of the Irish Centre for High-End Computing said high-perfromance computing has become a fundamental part of research and scientific advancement across all disciplines, particularly in areas such as climate informatics, digital twins of Earth systems, healthcare and medicine, and material sciences.
10 things you need to know direct to your inbox every weekday. Sign up for the Daily Brief, Silicon Republic’s digest of essential sci-tech news.