China Intends to Exceed 300 Exaflops Aggregate Compute Power by 2025

Print Friendly, PDF & Email

The People’s Republic of China, so cagy in recent years regarding its leadership-class supercomputing resources, on Sunday declared its intent to exceed 300 exaflops of aggregate computing power within two years, a 50 percent increase over its current stated capacity of 200 exaflops. This in the face of tightening sanctions imposed by the U.S. on the export of advanced microprocessors used for HPC and AI.

The “Action Plan for the High-Quality Development of Computing Infrastructure” is a joint statement issued by six government departments “proposing that by 2025, computing infrastructure will … (exceed) 300 EFLOPS, the proportion of intelligent computing power reaches 35 percent, and the eastern and western computing power develops in a balanced and coordinated manner.”

The announcement would tend to affirm a report last month in the South China Morning Post that a third Chinese exascale system may have come online, according to HPC community luminary Jack Dongarra, who recently visited the country. Another story, in The Financial Times earlier this year, reported that Dr. David Kahaner, director of the Asian Technology Information Program, said China is investing in 10 new exascale systems by 2025.

The new PRC statement offers only a broad picture of Chinese computing capacity goals, it does not provide information on the number of new supercomputers to be built, their architecture or other details. China has been famous the past five years for not submitting benchmark results to the Top500 list of the world’s most powerful supercomputers. But the country still has two top 10 systems, Sunway TaihuLight at no. 7 (joined list in 2016) and Tianhe-2A at no. 10 (joined list in 2018).

In its 300 exaflop statement, the PRC said the statement was jointly issued by six departments, including the Ministry of Industry and Information Technology, the Cyberspace Administration of China, the Ministry of Education, the National Health Commission, the People’s Bank of China, and the state-owned Assets Supervision and Administration Commission of the State Council.

The statement broke down China’s national compute infrastructure into three categories:

China’s “carrying” (network) capacity: “the national hub node data center clusters can basically achieve direct network transmission with a delay no higher than 1.5 times the theoretical delay. The optical transport network coverage rate in key application sites has reached 80 percent, and the backbone network and metropolitan area network are fully supported”

The country’s total storage capacity exceeds 1800EB, with advanced storage capacity accounting for more than 30 percent, with disaster recovery coverage rate of core data and important data in key industries approaching 100 percent.

Regarding “application empowerment,” the statement said “new computing power businesses, new models, and new formats have been created,” while the “penetration rate of computing power in industry, finance and other fields has been significantly increased.”