Chinese supercomputer manufacturer Sugon Information Industry Co. has unveiled its latest offering, Sugon 8000, an artificial intelligence (AI) supercluster built using domestically developed computing cards. The supercluster boasts 100,000 cards and has already been deployed to push new frontiers in different areas of research.
Even though the tech world is focused on developments in the fields of AI and how rapidly it will progress in the future, the bulk of new work is being done by supercomputers. Every now and then, a new AI model is released that perhaps can even solve mathematical problems that have puzzled us for centuries, but supercomputing infrastructure is still the first choice for researchers working on problems.
An AI supercluster offers the dual advantage of supercomputing infrastructure, while also adding elements of AI that bring the best of both worlds when solving a problem.
Beijing-headquartered Sugon Information Industry Co. is a spinoff from the Chinese Academy of Sciences and launched its first supercomputer in 1993. Since then, it has developed multiple supercomputers and recently unveiled Sugon 8000 at the China National Annual Conference on High Performance Computing.
100,000-card AI supercluster
The Sugon 8000 features 100,000 computing cards interconnected to form a major computing cluster, a major feat in computing hardware. What makes it even more special is that China built all the computing cards domestically.
The push to merge high-performance computing (HPC) infrastructure with AI lets users allocate computing resources based be the needs of different tasks. The Sugon 8000 is connected to the National Supercomputing Internet, China’s national integrated computing network, which makes it accessible to the industry, research institutes, individuals, as well as private enterprises in the country.
The scale of the system also signifies China’s capacity to build high-end computing infrastructure locally. Putting together a system like Sugon 8000 is not just about aggregating chips but also managing high-speed interconnects, liquid cooling, work scheduling, and long-term stability of the system, all of which were achieved domestically.
This is a major achievement for China as it works its way from manufacturing HPC infrastructure to deploying it for practical use for research and industrial applications.
New frontiers of research
Prior to the unveiling at the Conference, Sugon 8000 had already been deployed for over 10 research applications in broad areas from materials to celestial bodies, aerospace to agriculture, energy and meteorology.
According to a local media report, multiple studies that used the Sugon 8000 made major breakthroughs in their research, showing that the AI supercluster can be extremely helpful in advancing new frontiers in research in the near future.
The developer team has optimized more than 300 key applications across 30 research and industrial fields for the Sugon 8000 system. As AI infiltrates more sectors, Sugon is being prepared to support scientific simulations, data analysis, and model training in the future.
Developing such AI superclusters locally will also help China cut research and development costs, shorten innovation cycles, and improve efficiency.