Frontier Still King: Exascale Supercomputer Tops Global List Again

Recently, the 63rd edition of the Global Supercomputer Top500 list was released, with the Oak Ridge National Laboratory’s Frontier, the world’s first supercomputer to reach the ExaFLOP level, retaining its top position. Its performance in the high-performance Linpack (HPL) benchmark reached 1.206 Exaflop/s, a slight improvement from before.

For a long time, Frontier has been the sole supercomputer to achieve ExaFLOP performance. It boasts a combination of 8,699,904 CPU and GPU cores, with the HPE Cray EX architecture integrating AMD’s third-generation EPYC CPUs and Instinct MI250X accelerators, optimized for high-performance computing (HPC) and artificial intelligence (AI). Data transfer between components is facilitated by Cray’s Slingshot 11 network. Notably, Frontier has a power efficiency rating of 52.93 GFlops/Watt, ranking 13th in the GREEN500, showcasing its exceptional efficiency.

Aurora, built on Intel‘s Sapphire Rapids and Ponte Vecchio, secured second place. This supercomputer, located at the U.S. Department of Energy’s Argonne National Laboratory, has faced multiple delays and is not yet fully operational. Nevertheless, it has become the second supercomputer to reach the ExaFLOP level, with performance increasing from 585.34 PetaFlop/s to 1.012 Exaflop/s.

Microsoft Azure Cloud’s Eagle ranked third, making it the highest-ranked cloud service system on the Top500 list, with a performance of 561.2 PetaFlop/s. It is built on Intel Xeon Platinum 8480C processors and Nvidia H100 accelerators. Japan’s Fugaku, which held the top position from June 2020 to November 2021, maintained its fourth place with a performance of 442 PetaFlop/s. Finland’s EuroHPC/CSC LUMI ranked fifth with a performance of 380 PetaFlop/s, making it the highest-ranked system in Europe.

The only new entrant in the top ten is the Alps system from the Swiss National Supercomputing Centre (CSCS), which ranked sixth with a performance of 270 PetaFlop/s.