For the first time since 2017, a Chinese supercomputer has claimed the world’s top spot on the TOP500 list. The LineShine machine in Shenzhen achieved this by running solely on domestic CPUs and no American chips, signaling a significant milestone amid ongoing US export restrictions.
- LineShine clocked 2.198 exaflops, beating US leader El Capitan by 20%
- Fully CPU-based design avoids Nvidia, AMD, Intel components amid export curbs
- China’s custom Arm processors power advanced simulations with high efficiency
What happened
China’s LineShine supercomputer, housed at the National Supercomputing Center in Shenzhen, was declared the world’s fastest at the ISC conference in Hamburg. It achieved 2.198 exaflops on the high-precision TOP500 benchmark, topping the previous leader El Capitan by more than 20%. This marks the first time since 2017 that a Chinese system has held the number one rank.
The defining feature of LineShine is that it operates without any US-sourced chips from Nvidia, AMD, or Intel. Instead, it uses nearly 14 million custom Arm-based CPU cores in 90 cabinets, leveraging a Chinese-developed software stack including KylinOS and a proprietary network. This makes it the first system to achieve sustained double-precision exascale performance using CPUs alone.
Why it matters
LineShine’s achievement underscores China’s successful push for technological self-reliance in supercomputing amid escalating US export controls. By avoiding American GPUs and CPUs, China demonstrates an advanced domestic semiconductor capability that can compete at the highest levels of scientific computing without reliance on Western suppliers.
This breakthrough exposes gaps in US export regulations, which focus primarily on restricting GPUs critical for AI workloads but apply less stringent controls on CPUs. Experts argue that these loopholes allow China to continue developing powerful systems like LineShine, challenging US strategic objectives in technology leadership and national security.
What to watch next
The global supercomputing and semiconductor landscapes will be closely monitored to see if China scales production and deployment of similar CPU-driven systems. Further developments could accelerate the divergence of HPC designs away from GPU-accelerated architectures dominated by American firms like Nvidia and AMD.
Meanwhile, Washington may respond by tightening export controls on CPU technologies and manufacturing processes or expanding oversight over overseas subsidiaries. The progression of LineShine’s scientific applications and contest entries for top awards will also signal how China leverages this capability for advanced research and technology influence.