The 66th edition of the TOP500 list was announced at the SC25 conference in St Louis, Missouri. Organizers framed the update as evidence of continued U.S. leadership in high-performance computing, while noting historic European milestones and growing global diversity across architectures and energy-efficient designs.
Lawrence Livermore National Laboratory’s El Capitan remains the undisputed leader. The system, built on the HPE Cray EX255a architecture and powered by AMD 4th Gen EPYC CPUs alongside AMD Instinct MI300A accelerators, recorded a remeasured HPL result of 1.809 exaflop per second. That remeasurement reinforces El Capitan’s position as the fastest supercomputer in the world.
El Capitan also leads across other benchmarks that target different performance characteristics. It posted 17.41 HPCG-Petaflop per second on the HPCG benchmark, which focuses on real-world application performance, and achieved 16.7 exaflop per second on the HPL-MxP mixed-precision benchmark. The HPL-MxP result is presented as confirmation of the system’s versatility for Artificial Intelligence and data-driven workloads.
The system’s scale and efficiency were highlighted as part of its overall profile. El Capitan comprises 11.34 million cores and delivers an energy efficiency of 60.9 GFlops per watt, figures the report cites to illustrate the capabilities of the current generation of exascale computing platforms. Together, the benchmark and efficiency metrics underline how large-scale systems are being evaluated across raw throughput, application performance, and power consumption.
