Overview
- LineShine, which topped the TOP500 on Tuesday, recorded 2.198 FP64 exaflops on the Linpack benchmark and overtook the U.S. system El Capitan at 1.809 exaflops.
- The system is the first entry in the TOP500 to sustain more than 2 exaflops of double‑precision (FP64) performance using only CPUs, a milestone noted by TOP500 co‑founder Jack Dongarra.
- Submitters reported LineShine was built and deployed at the National Supercomputing Centre in Shenzhen using a domestic stack that includes semi‑custom LX2 (Armv9) CPUs, a LingQi interconnect and the Kylin OS, totaling about 13.79 million cores and drawing roughly 42.2 megawatts.
- Experts warn the Linpack FP64 result does not equal AI leadership because LineShine delivers weaker mixed‑precision and AI‑relevant performance, shows lower performance‑per‑watt than some rivals, and scored below GPU‑accelerated systems on mixed‑precision benchmarks.
- The result highlights China’s push for supply‑chain self‑reliance and sharpens US–China competition over advanced computing, a development likely to influence export controls, R&D priorities and national investment in accelerator‑based AI infrastructure.