China has deployed its first operational 30,000-card AI computing cluster at the Zhengzhou core node, marking a significant milestone in the country's domestic supercomputing infrastructure development.
China has put its first operational 30,000-card national supercomputing internet core node into trial operation, marking a major milestone in domestically built AI computing infrastructure. According to Sugon, on February 5, the National Supercomputing Internet Application Technology Conference and Core Node Trial Operation Launch Ceremony was held in Zhengzhou, Henan Province.
Three 10,000-card supercluster systems (scaleX 10,000-Card Super Cluster), provided by Sugon, simultaneously entered trial operation at the Zhengzhou core node of the National Supercomputing Internet. This marks the country's first operational deployment of a 30,000-card AI computing cluster, making it the largest domestically developed AI computing pool in actual operation to date.
The cluster fully supports large-scale AI workloads, including trillion-parameter model training, high-throughput inference, and AI for Science applications. The system's scale represents a significant achievement in China's push for computational self-reliance in AI development.
Background and Development
In 2023, China's Ministry of Science and Technology convened a dedicated meeting to coordinate and launch the development of the Supercomputing Internet. In April 2023, construction of the National Supercomputing Internet Platform officially began, with the establishment of the National Supercomputing Internet Consortium. The National High Performance Computing Engineering Technology Research Center serves as the consortium's chair unit.
On January 4, the National Supercomputing Internet Platform announced that its user base had officially exceeded one million users. The platform has now connected 14 provinces and municipalities, integrating more than 30 national supercomputing and intelligent computing centers. Together, they form the country's largest heterogeneous computing resource pool, with over 150,000 accelerator cards and more than 2 million CPU cores, offering users nearly 70 different computing configurations.
Technical Significance
The 30,000-card deployment represents a significant technical achievement in several areas:
- Scale: The cluster is the largest domestically developed AI computing pool currently in operation
- Integration: Three 10,000-card superclusters working in concert
- Application scope: Supports trillion-parameter model training and high-throughput inference
- Scientific computing: Enables AI for Science applications
The use of "scaleX 10,000-Card Super Cluster" architecture suggests a modular approach to building massive AI infrastructure, allowing for incremental scaling while maintaining system coherence.
Strategic Context
This deployment comes amid broader efforts by China to develop indigenous computing capabilities, reducing reliance on foreign technology in critical infrastructure. The National Supercomputing Internet represents a coordinated national effort to create a unified computational resource pool that can serve diverse needs across research, industry, and government sectors.
The achievement of one million users within a relatively short timeframe indicates rapid adoption and suggests the platform is addressing real computational needs across China's research and industrial ecosystem.
Looking Forward
As the trial operation phase progresses, key questions will include:
- How will the system's performance compare to international counterparts?
- What specific applications will drive initial usage?
- How will the infrastructure evolve to support emerging AI workloads?
- What role will this play in China's broader AI development strategy?
The successful deployment of this 30,000-card system represents not just a technical milestone but a statement of intent about China's position in the global AI infrastructure landscape.
Comments
Please log in or register to join the discussion