Test report DSG-5718 · Rev C · tested November 17, 2026
AI Datacenter InfrastructureDevice under test
LineShine holds Top500 crown as turnover hits record low
China's LineShine tops the November 2026 Top500 at 2.198 exaflops as new entrants fall to a record-low 27 and the top ten stays frozen for a second edition.
- Read
- 3 min
- Words
- 501
- Node
- 65nm
- Operator
- Priya Raman
Spec summary
- LineShine tops the November 2026 Top500 with 2.198 exaflops, ahead of El Capitan at 1.809 exaflops.
- Only 27 new systems entered the list, the lowest number ever recorded.
- LineShine uses 40,960 LX2 processors across 92 cabinets, totaling 13,789,440 cores.
- The highest new entrant, Hangang in South Korea, ranked 12th at 342.6 exaflops.
- The Green500 top ten is dominated by European systems: three in France, three in Germany, one each in the UK and Poland.

China's LineShine retained the top spot on the November 2026 Top500 list with 2.198 exaflops of HPL performance. The list recorded only 27 new entrants. The Top500 organizers called this "the lowest number ever recorded by a wide margin."
The top ten is unchanged for a second consecutive edition. LineShine took first place from the US El Capitan system in June 2026. El Capitan, at Lawrence Livermore National Laboratory, remains second at 1.809 exaflops.
New turnover has collapsed to a historic low. That figure invites scrutiny. The Top500 itself concedes the stagnation is not evidence of stalled progress. Most growth now comes from additions and upgrades near the top of the list, not from wholesale replacements of leading machines, the organizers said. Readers should note what the statement does not say: it does not explain why replacement activity has slowed.
The highest-ranking newcomer did not reach the top ten. The 342.6 exaflops Hangang system, installed at the KITSI Supercomputing Center in South Korea, entered at 12th place. It runs on Nvidia Grace CPUs alongside GH200 Superchips (combined CPU-GPU modules marketed by Nvidia).
What sits inside LineShine?
The system is housed at the National Supercomputing Centre in Shenzhen. It uses Chinese-designed chips and networking gear throughout. Its processors are semi-custom 304-core LX2 parts based on the Armv9 instruction set architecture (the instruction set licensed from Arm) running at 1.55GHz.
The LX2 appears to have been co-designed with China's National Supercomputing Center and Huawei. The word "appears" is the Top500's own hedge. The list does not confirm the partnership outright. Some 40,960 chips are deployed across 92 cabinets, giving a total of 13,789,440 cores. For comparison, that core count is roughly ten times the 1.4 million cores of El Capitan's predecessor-class designs at similar rankings a few years ago.
How does the Green500 compare?
The Green500 list of most energy-efficient supercomputers also shows an unchanged top ten. Its composition differs sharply from the raw performance ranking. Europe dominates. Three systems sit in France, three in Germany, and one each in the United Kingdom and Poland. The traditional Top500 top ten instead holds four US entrants alongside systems from China and Japan.
The full HPL top ten
- LineShine, National Supercomputer Center, Shenzhen, China — 2.198 exaflops
- El Capitan, Lawrence Livermore National Laboratory, United States — 1.809 exaflops
- Frontier, Oak Ridge National Laboratory, United States — 1.353 exaflops
- Aurora, Argonne National Laboratory, United States — 1.012 exaflops
- JUPITER Booster, Jülich Supercomputing Centre, Germany — 1.000 exaflops
- HPC7, Eni, Italy — 571.5 petaflops
- Eagle, Microsoft, United States — 561.2 petaflops
- HPC6, Eni, Italy — 477.9 petaflops
- Fugaku, RIKEN Center for Computational Science, Japan — 442.01 petaflops
- Alps, Swiss National Supercomputing Centre (CSCS), Switzerland — 434.9 petaflops
Two editions without a change at the top is unusual. Three would establish a pattern. The June 2027 list will show whether upgrades, not replacements, can keep moving the frontier.
via datacenterdynamics.com (Source)
More from Priya Raman
Show full bio
Correspondent covering money, funding and market shifts around hardware layer of artificial intelligence at Die Signal.
661 articles
Same lot · LOT-362D
- DSG-953028nmKorea's Hangang Supercomputer Debuts 12th on TOP500, Missing Top-10 Goal
- DSG-433414nmChina's LineShine Tops June 2026 Top500 at 2.74 Exaflops Peak
- DSG-210565nmChina's LineShine Tops Top500 with 2.2 Exaflops All-CPU Architecture
- DSG-960210nmSMIC Posts 20% Surge, Takes Third Place From GlobalFoundries
- DSG-185010nmDo We Still Need GPUs? Three HPC Researchers Say Maybe Not