Second-place Frontier at the Oak Ridge National Laboratory clocked an HPL score of 1.353 EFlop/s, which is same as in June. Based on the HPE Cray EX235z architectures and equipped with AMD third-generation EPYC 64C 2GHz processors, the Frontier systems has 9,066,176 total cores and also relies on HPE Slingshot interconnect for data transfer.
The third U.S. system, Aurora, based at the Argonne Leadership Computing Facility, was submitted with 1.012 EFlop/s on the HPL benchmark, which is unchanged from June. Built by Intel and based on the HPE Cray EX-Intel Exascale Compute Blade, the Aurora system communicates through HPE Slingshot interconnect and uses Intel Xeon CPU Max Series processors.
In fourth place on the list is JUPITER (JU Pioneer for Innovative and Transformative Exascale Research) at the EuroHPC/Jülich Supercomputing Centre in Germany. It was being commissioned at the time of the last listing and made a preliminary HPL value of 793.4 Petaflop/s on a partial system. Now fully deployed, it has achieved a benchmark of exactly 1 Exaflop performance. It is built on an Eviden’s BullSequana XH3000 direct liquid-cooled architecture using Nvidia Grace Hopper GH200 chips with a total of 4,801,344 cores.
