The first highly efficient distributed GPU superco

  • Detail

China's first high-performance distributed GPU supercomputing system launched Beijing, April 24 China's first distributed GPU supercomputing system was completed and officially launched in the Chinese Academy of Sciences on April 24, marking that China's ability to solve practical scientific research and engineering problems with high-performance computing has reached the international advanced level, and has entered the international leading ranks in terms of software and hardware coordination and energy efficiency

it is reported that high-performance computing plays an increasingly important role in scientific and technological progress and economic and social development in the world today. However, while the peak value of supercomputer is advancing by leaps and bounds, problems such as large investment, high operation and maintenance cost and low efficiency of data are becoming increasingly prominent. With the support of special funds from the Ministry of finance, the Chinese Academy of Sciences has successfully established a high-performance supercomputing system based on GPU, which must improve the displacement test accuracy to meet the user's requirements for times with a double precision peak value of more than 10 billion, and a single precision peak value of more than 3000 trillion times. In cooperation with Lenovo and Shuguang, the Chinese Academy of Sciences has promoted 10 sets of million billion times systems in research institutes in different fields within the Chinese Academy of Sciences, Thus, a distributed GPU supercomputing system with aggregate computing capacity of nearly 6000 trillion times is built. The system meets the needs of the majority of scientific and technological personnel for supercomputing, which is a combination of real variable 1 extensive plastic granulator and intelligent plastic granulator. It realizes efficient supercomputing with low cost and ready-made network facilities, and forms an application-oriented and efficiency first supercomputing mode with Chinese characteristics

at present, ten user units, including the Institute of process and the Institute of high energy physics of the Chinese Academy of Sciences and the National Astronomical Observatory, have used the industrial cluster system of high-end chemical new materials of related departments to carry out applications in high-energy physical simulation and experimental data analysis, astrophysics and space science research, petroleum geophysical data analysis and reservoir production process simulation, industrial and medical image processing, etc. The system is also serving national major projects, National Science and technology support plan, national major natural science funds and a number of large enterprises

it is understood that in order to promote the independent research and development of China's scientific research equipment, the Ministry of Finance arranged special funds to carry out the pilot work of independent research and development of national major scientific research equipment in the Chinese Academy of Sciences during the Eleventh Five Year Plan period. During the development of this supercomputing system, the Ministry of Finance and the Chinese Academy of Sciences jointly explored the project management system and mechanism, and achieved remarkable results, providing a useful reference for the independent innovation of deploying scientific research equipment nationwide

Copyright © 2011 JIN SHI