Overview
- The 67th TOP500, presented at the ISC conference in Hamburg, lists LineShine at 2.198 exaflops on the HPL benchmark, putting it first and moving El Capitan to second place.
- LineShine ran its HPL result using 45,360 domestically made LX2 CPUs totaling about 13.8 million cores and a reported 2.736 exaflops theoretical peak performance.
- The system is CPU-only, uses a proprietary LingQi interconnect and Kylin OS, and drew roughly 42.2 megawatts during the Linpack run, highlighting trade-offs between raw HPL speed and energy cost.
- LineShine also tops the memory- and network‑bound HPCG metric at about 22 petaflops but ranks fourth on the accelerator‑optimized HPL‑MxP metric (7.92 exaflops), showing different architectures favor different workloads.
- The result marks the first time since 2017 a Chinese machine leads the TOP500 and underscores China’s push for hardware self‑reliance after export controls, while Europe’s Jupiter Booster remains the fastest system on the continent.