At the top of the performance ladder, El Capitan, located at Lawrence Livermore National Laboratory in California, holds onto its title as the fastest supercomputer on Earth. Built on the HPE Cray EX255a platform, it delivers an astonishing 1.742 exaflops of computing power, thanks to AMD’s latest 4th-gen EPYC CPUs and MI300A accelerators. With over 11 million cores and top-tier energy efficiency, it’s setting new standards in high-performance computing (HPC).
Right behind it, Frontier, the long-time leader housed at Oak Ridge National Laboratory, continues to impress with 1.353 exaflops. This system, built on a previous-generation AMD architecture, still offers one of the most balanced combinations of raw performance and scalability, with nearly 8.7 million cores connected via the HPE Slingshot interconnect.
Taking third place is Aurora, Intel’s flagship exascale system at the Argonne Leadership Computing Facility in Illinois. With over 9.2 million cores and 1.012 exaflops of compute power, Aurora showcases the capabilities of Intel’s Xeon Max Series chips in a high-density, HPE-based architecture.
A strong new contender from Europe, the JUPITER Booster system at Forschungszentrum Jülich in Germany, is already making waves. Though still in the commissioning phase, a partial deployment has achieved 793.4 petaflops, making it the most powerful system in Europe to date. It runs on GH Superchips within Eviden’s BullSequana XH3000 design, cooled by a direct liquid architecture and supported by Nvidia InfiniBand.
Other top-ranking systems include:
- Eagle, a powerful Microsoft Azure cloud-based supercomputer, delivering over 560 petaflops using Intel Xeon Platinum chips and Nvidia networking.
- HPC6 in Italy, deployed at the Green Data Center by Eni, with nearly 478 petaflops and optimized for energy-intensive simulations using AMD EPYC and MI250X accelerators.
- Japan’s Fugaku, once the world’s fastest, still holding strong with over 442 petaflops, built on Fujitsu’s proprietary A64FX ARM-based architecture.
- Switzerland’s Alps, a high-efficiency system powered by Nvidia Grace and GH200 Superchips, pushing 435 petaflops.
- Finland’s LUMI, part of the EuroHPC initiative, with a performance of 379 petaflops, running on AMD’s 3rd-gen EPYC.
- And Leonardo, based in Italy, built with Intel Xeon processors and Nvidia InfiniBand, rounding out the top ten with 241 petaflops.