The United States regained the top spot in the field of supercomputers in the Top500 ranking of the most powerful systems from Japan.
Oak Ridge National Laboratory’s Frontier system powered by AMD EPYC CPUs took first place from last year’s champion, Japan’s Fugaku system powered by ARM A64X CPUs.
The Frontier system is still in the process of integration and testing at Oak Ridge National Laboratory in Tennessee. But it is operated by the US Air Force and the US Department of Energy.
Powered by Hewlett-Packard’s Cray EX235a platform and equipped with a 2GHz AMD EPYC 64C processor, the Frontier was also the top device by a large margin.
The new Cray EX platform combines third-generation AMD EPYC CPUs optimized for HPC and AI with AMD Instinct 250X accelerators and Slingshot-11 connectivity.
Frontier represents the first true exascale system, reaching a peak of 1.102 exaflops on the Lynmark scale. The system contains a total of 8,731,112 cores, and relies on Gigabit Ethernet for data transmission.
Meanwhile, the RIKEN Center for Computational Science’s Fugaku made less than half that at 442 petaflops. And that was enough to keep him at number one for the past two years.
Due to the fact that the theoretical fugaku peak is above the 1 exaflop barrier, there is also reason to call this system an exascale as well. However, Frontier is the only system capable of demonstrating this in a standard HPL test.
The Frontier was also the most efficient supercomputer. It operates at 52.23 gigaflops per watt, and surpassed the Japanese MN-3 system to occupy the first place in the Green500 list.
“It is amazing that the world’s fastest machine is also the most energy efficient,” said the director of Oak Ridge National Laboratory.
America regains the first place in the field of supercomputers
Other hardware in the TOP10 includes the LUMI system at EUROHPC in Finland based on Hewlett Packard’s Cray EX platform.
This new system is now ranked third. It contains 1,110,144 cores and has an HPL of approximately 151.9 petaflops. It is also worth noting that LUMI is the largest system in Europe.
The Adastra system at GENCI-CINES in France is also a new addition. The third new HPE Cray EX system achieved an HPL reference score of 46.1 petaflops, and is the second most powerful machine in Europe, after LUMI.
There’s the IBM-built Summit with 22-core CPUs and NVIDIA Tesla V100 GPUs (148.8 petaflops) and Sierra, a smaller version of Summit at 94.6 petaflops.
China ranked sixth and ninth with Sunway TaihuLight and Tianhe-2A. Both systems are located at the National Research Center for Parallel Computer Engineering and Technology and China National University of Defense Technology.
However, China is rumored to have at least two exascale systems (as per the Lynmark standard) across the new Sunway Oceanlite and Tianhe-3 systems. But due to the current state of its semiconductor policies, China is not revealing any new standards or significant developments.