New Delhi: China’s LineShine is the world’s fastest supercomputer. It was developed at the National Supercomputing Center in Shenzhen, has topped the latest TOP500 rankings, outperforming its closest rival in the United States.
Released twice a year, the TOP500 ranks the world’s most powerful supercomputers using the Linpack benchmark, which measures how quickly they solve complex mathematical calculations. LineShine achieved more than 2.19 exaflops, or over two quintillion calculations every second, making it about 22 percent faster than El Capitan, the US-based system at Lawrence Livermore National Laboratory that previously held the top spot.
Unlike most modern supercomputers, LineShine does not rely on separate graphics processing units (GPUs), which have become essential for artificial intelligence workloads. Instead, the Chinese system combines traditional computing and AI capabilities within the same architecture using nearly 14 million processing cores and specialised computing units.
The achievement is particularly significant because the machine has been built largely with Chinese technology at a time when the United States has imposed export restrictions on advanced AI chips to China. Researchers say the system demonstrates China’s growing ability to develop high-performance computing hardware despite those curbs.
Also read: How an NPR podcast helped prove that an hourly ‘movement break’ is good for you
Next generation of scientific computing
According to computer scientist Jack Dongarra of the University of Tennessee, who co-created the TOP500 rankings, LineShine represents an important step in bringing together conventional scientific computing and AI on a single platform. As scientific research increasingly incorporates machine learning, systems that can efficiently handle both types of workloads are expected to become more valuable.
Researchers have already begun using the supercomputer for real-world scientific problems. Haohuan Fu of Tsinghua University said his team used LineShine to combine traditional weather models with AI to improve rainfall forecasts across East Asia. The machine processed weather predictions spanning 2016 to 2025 in just 14.6 hours, a task that would be difficult for conventional computing clusters. Scientists have also used it to simulate magnetic materials at the atomic level.
The computer’s chips are partly based on designs originally developed by British semiconductor company Arm but include Chinese innovations, including built-in hardware for AI calculations and stacked memory placed close to each processing core to speed up data transfer. By integrating functions traditionally handled separately by CPUs and GPUs, the distinction between the two types of processors is becoming increasingly blurred.
Chinese officials have described the milestone as evidence of the country’s growing technological self-reliance. Analysts say LineShine’s success also highlights how export restrictions may have accelerated domestic innovation in China’s supercomputing sector.