The International Supercomputer Conference in Hamburg was the perfect platform for Lineshine (or Língshèng 灵晟) to, well, shine, as it received the crown of the world’s most powerful computer ever built.
It smashed the previous TOP500 record held by El Capitan, the US supercomputer built in 2024 by HPE for the National Nuclear Security Administration (NNSA) at the Lawrence Livermore National Laboratory (LLNL).
Unlike the latter, Lineshine uses CPU power only rather than APU or GPUs, a remarkable feat given the prevalence of these technologies. Only another supercomputer in the top 10, the Arm-powered Fugaku supercomputer, falls in that category.
Details shared by Lu Yutong, the chief designer of Lineshine show that it is powered by more than 13.7 million ARMv9 cores spread over 90 racks and 45,360 CPUs.
Yes, you’ve got it right, there are 304 cores per socket split in two chiplets, and eight NUMA domains each with 38 cores and 4GB HBM memory. There’s apparently eight CPU slots left empty per rack.
Each socket can access 256GB of DDR5 memory, which means that the entire system has 128TB off-package RAM and 16TB HBM. The entire supercomputer has 11.6PB of traditional DDR5 memory and almost 1.5PB of HBM.
Compared to El Capitan pure HBM3 approach, Lineshine opted for a tiered memory pool similar to traditional computers (using RAM and SSD). The system is connected to 200PB of direct storage as well.
Breathtaking performance
Altogether, Lineshine reached 2.198 exaflops, 21% faster than the 1.809 exaflop figure reached by El Capitan.
However numbers don’t tell the whole story; China’s new top dog consumes far more power than El Capitan, making it far less efficient. It only reached 52 Gigaflops/Watt compared to 60.95 Gigaflops for El Capitan and 73.28 for Kairos, the greenest supercomputer in operation.
What is more interesting is the fact that each core, even running at 1.5GHz, reached about 200 Gigaflops of FP64. Significantly more than any supercomputing CPU Core (e.g. AMD Zen 4 or Nvidia Grace) which ironically underlines Chinese weakness when it comes to GPU, i.e they went the CPU way because they didn’t have any option than to go brute-force.
And that’s despite the barrage of launches by Huawei over the past few years with the Ascend 910D, the Ascend 910C and the Ascend 920.
Lineshine used its own version of Nvidia’s Nvlink interconnect called LinQi, one that can scale in excess of 100,000 nodes or more than 60 million cores, 4x the current core count. So there’s plenty of room to grow should Lineshine want to keep its rank.
It is not the first time that China sits atop the TOP500 leaderboard. The last time it achieved it was with the Sunway TaihuLight supercomputer back in 2017. And while nabbing the top spot on this leaderboard gets the winner bragging rights, it is only really useful for HPC (high performance computing) applications like CFD, earthquake simulation, materials, energy, drug design, neuroscience and scientific AI, and others.
Hyperscalers like Google or Microsoft can muster even more powerful compute power, by at least an order of magnitude, if needed but that is not accounted for by TOP500 for various reasons.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.