Chinese computer researchers using domestically made graphics processors have achieved nearly three boosts in performance over the powerful US supercomputers they rely on nvidiaPeer-reviewed research shows cutting edge hardware.
This achievement identifies the potential unintended consequences of escalating technical sanctions in Washington.
Researchers said innovative software optimization techniques have allowed them to improve the efficiency of computers designed in China. Graphic Processing Unit (GPU) surpasses US supercomputers in certain scientific calculations.
Skeptics warn that software tweaks alone cannot fill hardware gaps indefinitely, but they emphasize that development highlights Beijing’s broader strategy.Chalk Point“The risks of important technology.
Scientists often rely on simulations to model real-world situations, such as designing to protect against flooding and urban flooding. However, such replication, especially large-scale, high-resolution simulations, require considerable time and computational resources, limiting the broader applications of such approaches.
The challenges for Chinese scientists are even more challenging. For hardware, advanced GPU production such as: A100 The H100 is controlled by foreign manufacturers. On the software side, US-based Nvidia has restricted it CUDA Software Ecosystem Running on third-party hardware prevents the development of independent algorithms.
In search of a breakthrough, Professor Nan Tongchau, who has a major laboratory in the state of hydrological water resources and hydroelectric engineering at Hohai University in Nanjing, has announced that “multi-node, multi-GPU” parallel computers based on domestic CPUs and GPUs. We have begun researching the fusion approach. The results of their study were published in the Chinese Journal of Hydraulic Engineering on January 3rd.
