The International Supercomputer Conference in Hamburg was the perfect platform for Lineshine (or Língshèng 灵晟) to shine, as it was awarded the crown of the most powerful computer ever built in the world.
It shattered the previous TOP500 record held by El Capitan, the American 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 only CPU power rather than APUs or GPUs, a remarkable feat considering the prevalence of these technologies. Only one other supercomputer in the top 10, the Arm-powered Fugaku supercomputer, falls into this category.
Details shared by Lu Yutong, Lineshine’s chief designer, show that it is powered by more than 13.7 million ARMv9 cores spread across 90 racks and 45,360 processors.
Yes, you’re right, there are 304 cores per socket split into two chipsets and eight NUMA domains each with 38 cores and 4GB of HBM memory. There are apparently eight empty CPU slots per rack.
Each socket can access 256GB of DDR5 memory, meaning the entire system has 128TB of out-of-package RAM and 16TB of HBM. The entire supercomputer has 11.6 PB of traditional DDR5 memory and almost 1.5 PB of HBM.
Compared to El Capitan’s pure HBM3 approach, Lineshine opted for a tiered memory pool similar to traditional computers (using RAM and SSD). The system is also connected to 200 PB of direct storage.
Breathtaking performance
In total, Lineshine reached 2,198 exaflops, 21% faster than the 1,809 exaflops achieved by El Capitan.
However, the numbers don’t tell the whole story; China’s new leader uses much more power than El Capitan, making it much 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’s more interesting is the fact that each core, even running at 1.5 GHz, achieves around 200 gigaflops of FP64. Much more than any supercomputing CPU core (e.g. AMD Zen 4 or Nvidia Grace), which ironically highlights China’s GPU weakness, i.e. they chose the CPU route because they had no choice but to go for brute force.
And this, despite Huawei’s numerous launches in recent years with the Ascend 910D, Ascend 910C and Ascend 920.
Lineshine used its own version of Nvidia’s Nvlink interconnect called LinQi, capable of scaling to over 100,000 nodes or over 60 million cores, or 4 times the current core count. So there is plenty of room for growth if Lineshine wants to maintain its rank.
This is not the first time that China has ranked first in the TOP500 rankings. The last time it achieved this was with the Sunway supercomputer TaihuLight, in 2017. And while landing the top spot in this ranking gives the winner bragging rights, it is only really useful for HPC (high performance computing) applications such as CFD, earthquake simulation, materials, energy, drug design, neuroscience and scientific AI, among others.
Hyperscalers like Google or Microsoft can muster even more powerful computing power, at least by an order of magnitude, if necessary, but this is not taken into account by the TOP500 for various reasons.
Follow TechRadar on Google News And add us as your favorite source to get our news, reviews and expert opinions in your feeds.
