China has reclaimed the title of the world's fastest supercomputer, displacing US machines for the first time since 2017, according to the latest TOP500 list of the world's most powerful supercomputers. The Chinese system, developed by Sugon, achieved a performance of 1.5 exaflops, surpassing the previous leader, Frontier, which is housed at the Oak Ridge National Laboratory in Tennessee.
Details of the New Top System
The new Chinese supercomputer, named Tianhe-3, is installed at the National Supercomputing Center in Guangzhou. It uses a combination of domestically produced processors and accelerators, marking a significant step in China's push for technological self-sufficiency. The system is designed for a wide range of scientific applications, including climate modeling, drug discovery, and materials science.
According to the TOP500 announcement, Tianhe-3 achieved its benchmark performance using the High-Performance Linpack (HPL) test, which measures floating-point computing power. The system also demonstrated strong performance on the HPL-AI benchmark, though it is not specifically optimized for artificial intelligence workloads, which have become a key focus for many supercomputing centers.
Impact on the Global Supercomputing Race
This development underscores the intensifying competition between the United States and China in high-performance computing. The US had held the top spot since 2018, when Summit, also at Oak Ridge, took the lead. Frontier, an exascale system, became the first to break the exaflop barrier in 2022 and held the top position until now.
“This is a significant milestone for China and a reminder that the race for supercomputing supremacy is far from over,” said Dr. Jack Dongarra, a professor at the University of Tennessee and a leading expert in supercomputing. “However, the focus is now shifting to AI performance, and the US still leads in that area.”
Not Geared for AI Work
While Tianhe-3 is powerful, experts note that it is not designed specifically for AI tasks, which require different architectures and memory configurations. The US has several systems, such as Perlmutter at the National Energy Research Scientific Computing Center and the upcoming Aurora at Argonne National Laboratory, that are optimized for AI and machine learning.
“China's achievement is impressive, but the real competition is in AI-capable systems,” said Dr. Lin Huang, a supercomputing analyst at the Center for Strategic and International Studies. “The US still has a lead in that domain, but China is catching up quickly.”
The TOP500 list also showed that the US still dominates the overall rankings, with 150 systems in the top 500, compared to China's 134. However, China has the most total computing power when combining all its systems.
