{"id":132348,"date":"2026-06-24T04:54:58","date_gmt":"2026-06-24T04:54:58","guid":{"rendered":"https:\/\/mailinvest.blog\/index.php\/2026\/06\/24\/chinese-supercomputer-using-local-processors-heads-top500-list\/"},"modified":"2026-06-24T04:56:03","modified_gmt":"2026-06-24T04:56:03","slug":"chinese-supercomputer-using-local-processors-heads-top500-list","status":"publish","type":"post","link":"https:\/\/mailinvest.blog\/index.php\/2026\/06\/24\/chinese-supercomputer-using-local-processors-heads-top500-list\/","title":{"rendered":"Chinese supercomputer using local processors heads TOP500 list"},"content":{"rendered":"

<\/a>\n
<\/p>\n

\n

HPC<\/p>\n

Use of Arm cores and Linux imply Beijing hasn\u2019t damaged away from the world<\/p>\n<\/div>\n

\n

The TOP500 record of Earth\u2019s mightiest supercomputers has a brand new chief: the two.198 Exaflop\/s LineShine machine housed on the Nationwide Supercomputer Heart (NSC) in Shenzhen, which took the highest spot with out utilizing any equipment from Nvidia, Intel, or AMD.<\/p>\n

Which isn’t to say that LineShine is a completely Chinese language creation. As defined in a pre-press paper<\/a>, the machine\u2019s LX2 processors are a neighborhood effort however use Armv9 designs \u2013 so chalk up a win for Blighty, the house of Arm. The machine additionally runs KylinOS \u2013 a Linux distribution<\/a> that options contributions from around the globe.<\/p>\n

The paper reveals that LineShine includes 20,480 computing nodes, and that every LX2 processor \u201cintegrates two compute dies (304 cores whole) and eight on-package HBM stacks (32 GB, 4 TB\/s mixture bandwidth).\u201d<\/p>\n

\u201cEvery compute die accommodates 152 cores and 128 GB of off-package DDR reminiscence organized into 4 NUMA domains,\u201d the paper provides. \u201cA devoted SDMA engine handles information motion between DDR and HBM. The LX2 helps FP64\/FP32\/FP16\/INT8 by way of SME and SVE items, delivering as much as 60.3\/120.6 TFLOPS in FP64\/FP32. Nodes are interconnected by way of the LingQi high-speed community with a dual-plane multi-rail fat-tree topology, providing 1.6 Tb\/s bandwidth per node.\u201d<\/p>\n

That community can also be a Chinese language creation, from the minds at Hangzhou LingQi Know-how Co.<\/p>\n

LineShine grew to become the primary system on the TOP500 to exceed two exaflops of sustained double-precision efficiency utilizing CPUs solely and the curators of the record suppose it might do higher in future assessments, as a result of this time round it reached about 80 % of its 2.736 Exaflop\/s theoretical peak in assessments carried out in preparation for this iteration of the TOP500 record.<\/p>\n

Information that LineShine topped the supercomputing charts comes as China\u2019s authorities more and more steers native organizations in direction of shopping for made-in-China tech. Beijing needs to lower dependency on overseas merchandise, as a result of China has gone all-in on AI and different applied sciences to spice up financial progress and improve the capabilities of its army.<\/p>\n

China\u2019s Communist Social gathering understands that reliance on imports can stymie these ambitions, with the USA\u2019s ban on GPU gross sales to the Center Kingdom providing ample proof of the necessity to management tech provide chains. And now Beijing can level to its insurance policies producing probably the most highly effective single pc on the planet.<\/p>\n

It\u2019s conceivable that China might do even higher sooner or later, as its GPU business is nascent and at present producing merchandise whose efficiency trails Nvidia and AMD by 4 or 5 years.<\/p>\n

These two paragons of US computing energy, together with Intel, dominate this model of the TOP500 record \u2013 as has been the case for years. China is subsequently on the march, however is a good distance from international dominance.<\/p>\n

Our sibling web site The Subsequent Platform has in depth evaluation of the TOP500 record here<\/a>. \u00ae<\/p>\n

\u00a0<\/p>\n<\/div>\n