China’s largest artificial intelligence computing cluster for scientific research entered operation on Tuesday, doubling its number of domestically made AI accelerator chips in just two months, according to state broadcaster CCTV.
The AI acceleration cards were produced by Chinese supercomputer developer Sugon, which is affiliated with the Chinese Academy of Sciences, and are running in the core node of the national supercomputing network in Zhengzhou, Henan province.
The number of chips in the computing node reached 60,000 units, up from 30,000 when trial operations began in early February.
With the upgrade, the Zhengzhou core node had become the country’s most powerful scientific intelligent computing infrastructure, CCTV reported.
CCTV said the development was “a breakthrough for China in computing infrastructure for AI-driven scientific research, which will help the country seize the commanding heights of AI industrial applications”.
Chinese researchers have long faced challenges in conducting “AI for science” research, including a shortage of computing power, software limitations and reliance on foreign suppliers for important tools, according to the official Beijing Daily.
The entire new “AI for science” computing cluster was developed domestically, from foundational chips to high-speed interconnects and an advanced software platform. It can also handle high-dimensional functions and complex scientific problems.
“This means that researchers can confidently run their most critical research tasks on this platform, no longer subject to external control,” Beijing Daily reported.
Last week, US lawmakers proposed a bill to further restrict China’s access to advanced chipmaking equipment by bringing allies such as the Netherlands and Japan to align with American curbs on semiconductor equipment exports to China within 150 days.
In November last year, the US also launched a national initiative called the Genesis Mission led by the Department of Energy and 17 national labs.
It sought to “build an integrated AI platform to harness federal scientific data sets ... to train scientific foundation models and create AI agents to test new hypotheses, automate research workflows and accelerate scientific breakthroughs”.
In the latest development in China, Sugon launched the country’s first one-stop development platform called OneScience. It hosts dozens of scientific models and data sets, and scientists do not need programming skills to use it.
Accelerated scientific progress can already be seen across multiple fields, according to Beijing Daily.
In drug development, the national Changping Laboratory focusing on life science has accelerated new drug creation by increasing the simulation speed of protein folding by three to six orders of magnitude.
Major drug discovery calculations that once required years of work can now be completed in just days.
In materials science, the Chinese Academy of Sciences’ Institute of Computing Technology has likewise reduced material screening time from years to days.
The cluster also supports research in new materials, brain science and aerospace by enabling a massive simulation of 41.47 billion atoms. It completed the first simulation of the human brain’s 86 billion neurons and performed turbulence simulations in trillions of grids. -- SOUTH CHINA MORNING POST
