• Date:
  • Category: Tech

On June 23, 2026, the "Lingsheng" (LineShine) supercomputer was ranked No. 1 on the global TOP500 list at the ISC 2026 conference in Hamburg, Germany, achieving a sustained double-precision performance of 2.19 EFLOPS (quintillion floating-point operations per second). It is the world's first supercomputer to exceed 2 EFLOPS in sustained performance and marks China's return to the top of the list for the first time in nine years, since the Sunway TaihuLight in 2017.

Technical Architecture

Lingsheng adopts a pure CPU design without relying on GPU accelerators, a notable departure from the mainstream CPU-GPU heterogeneous architecture. The system is based on the domestically developed "Lingkun" platform and features:

  • LX2 Processors: 304 cores per processor, based on the ARMv9 architecture, running at 1.55GHz, with on-chip matrix acceleration units for AI and scientific computing.
  • System Scale: 20,480 compute nodes, each with 2 LX2 processors, totaling nearly 14 million CPU cores across the entire system.
  • Interconnect: The domestic "Lingqi" high-speed network provides 1.6 Tbps bandwidth per node and supports up to 2 million ports.
  • Cooling: The system features the industry's first 100% liquid-cooled cabinets with an energy efficiency ratio of 51 GFlops/W, setting a new benchmark for green computing.
  • Operating System: Runs the Kylin operating system.

Significance and Application

The Lingsheng system is a fully domestic full-stack infrastructure designed for scientific and engineering intelligent computing, covering chips, networking, storage, system software, and cooling. It has already been applied in multiple fields, including atmospheric and ocean modeling, materials science, drug discovery, brain science, scientific AI, and large-scale model inference, achieving an average parallel efficiency of 84.4%.

The system was designed by Professor Lu Yutong, Director of the National Supercomputing Center in Shenzhen and Professor at Sun Yat-sen University. It has earned high praise from international experts, including Turing Award winner Jack Dongarra, who called it "a beacon of hope for the new architecture of AI4Science".

Lingsheng's achievement demonstrates that an all-CPU architecture can compete at the highest level, validating an alternative pathway in high-performance computing without relying on external GPUs, which are subject to export restrictions.