Domestic GPUs outperform Nvidia chips by nearly 10 times in supercomputer simulations, Chinese scholars say.

In January, Nan Tongchao, a professor at Hohai University's Yangtze Institute for Conservation and Development, led a team to release a peer-reviewed study.

This study is based on the "multi-node-multi-GPU" large-scale distributed heterogeneous parallel computing hydrodynamic model developed on a domestic hardware platform, which is used for efficient parallel solution of two-dimensional shallow water equations at the basin scale.

The results show that in supercomputer simulations, the performance of China's domestic graphics processing unit (GPU) exceeds that of Nvidia chips by nearly 10 times.

Hong Kong's South China Morning Post previously reported that this result shows that the U.S.' escalating technical sanctions on China are counterproductive and may not only have unintended consequences but also further challenge its dominance in manufacturing chips.

At present, this research report entitled "Heterogeneous Parallel Computing of Hydrodynamic Model Based on Domestic CPU/GPU Platform" has been made public in China's academic platform CNKI.net.

It is understood that the article will be published soon in the Chinese academic journal of water conservancy industry, Journal of Hydraulic Engineering.