LineShine, China’s New CPU-Only Exascale Supercomputer, Takes TOP500 Crown at ISC 2026
At ISC 2026 in Hamburg, China’s LineShine supercomputer claimed the TOP500’s top spot with 2.198 exaflops sustained FP64 performance. The system uses custom Armv9 CPUs and draws 42.22 megawatts, marking China’s first TOP500 appearance in nine years.
At the ISC High Performance 2026 conference in Hamburg, Germany, the 67th TOP500 list crowned a new number one: LineShine, a CPU-only supercomputer built in Shenzhen, China. This marks China’s first appearance on the TOP500 in nine years and its first top ranking since 2018.
LineShine achieved a sustained FP64 performance (Rmax) of 2.198 exaflops, just shy of its theoretical peak of 2.735 exaflops (Rpeak), according to the TOP500 listing. The system draws 42.22 megawatts of power, yielding a FP64 efficiency of 52.07 gigaflops per watt.
Custom LX2 CPU and massive core count
LineShine is powered by the LX2, a Chinese-designed Armv9-compliant CPU with SVE2 and SME support. Each LX2 package integrates two compute dies, each with four 40-core clusters, but two cores per cluster are disabled, leaving 38 active cores per cluster. That gives each die 152 active cores, and each LX2 totals 304 active cores running at 1.55 GHz, delivering 60.3 TFLOP/s of FP64 compute per chip at 690 watts.
The LX2 also includes eight stacks of on-package high-bandwidth memory totaling 32 GB with 4 TB/s bandwidth, and is backed by 256 GB of DDR5 memory per CPU. The full LineShine system comprises:
- Over 22,000 nodes, each with two LX2 CPUs and 800 Gbps of networking (1.6 Tbps per node)
- Eight nodes per compute blade, 16 blades per compute frame, and two frames per cabinet
- 90 compute cabinets, resulting in more than 13 million CPU cores
LineShine also tops HPCG, while Italy’s HPC7 enters top 10
LineShine did not stop at the LINPACK benchmark. It also achieved first place on the HPCG benchmark with 22.004 petaflops, surpassing the previous leader El Capitan’s 17.406 petaflops. This indicates the system is not a mere “LINPACK special” but excels at more realistic workloads.
Another notable entrant is the HPC7 system from Italy’s Eni, which landed at number six with 571.5 petaflops sustained FP64 using the same HPE Cray EX4000 platform and AMD Instinct MI300A APUs as El Capitan but at 30% of its size. Italy now has more compute on the TOP500 than any other European country.
The former number one, Fugaku, slipped to ninth place, but remains third in HPCG. Meanwhile, the Green500 top ten saw no changes for the first time in its history, though overall fleet efficiency improved as older systems were retired.
Future implications and handover to ACM SIGHPC
The appearance of LineShine raises questions about whether China will submit results for its other known exascale systems, Sunway Oceanlight and CNIS, which have remained off the list. It may also spur increased U.S. Department of Energy funding for larger supercomputers. Separately, the ISC Group announced it will hand over management of the TOP500 list to ACM SIGHPC, giving each future list a dedicated DOI for easier referencing.
Fact check
-
LineShine achieved an Rmax of 2.198 exaflops and an Rpeak of 2.735 exaflops.
reported · source
-
Each LX2 CPU has 304 active cores across two dies, with each die having 152 cores from four 40-core clusters (two cores disabled per cluster).
reported · source
-
LineShine is China’s first TOP500 submission in nine years and first number one since 2018.
reported · source
-
LineShine also topped the HPCG benchmark with 22.004 petaflops, beating El Capitan’s 17.406 petaflops.
reported · source
-
The ISC Group will hand over TOP500 list management to ACM SIGHPC, giving future lists DOI numbers.
reported · source
Source reporting (2)
Join the conversation
You need to be registered and logged in to comment on blog articles.
0 Comments
No comments yet
Be the first to share your thoughts on this article.