【#Tech24H】Recently, Sugon, together with a research team from Xi’an Jiaotong University, completed ultra-large-scale high-fidelity numerical simulation of multi-nozzle supersonic jets on China’s domestically produced 100,000-card AI supercluster Sugon 8000, setting the largest known record of its kind internationally. The results are aimed at research on reusable rockets for commercial spaceflight, providing high-precision computational means for structural safety assessment and thermal protection design. More noteworthy than the record is what lies behind it: Chinese computing power is moving from “being able to run” to “being easy to use”. Sugon 8000 is an AI supercluster built by Sugon based on a heterogeneous computing route. It adopts a heterogeneous computing architecture, balancing high-precision numerical simulation with AI training and inference, and providing a computational foundation for AI for Science (AI4S) research. At present, it has adapted to nearly 200 industry applications, covering more than 30 scientific research and industrial fields. In this collaboration, the team relied on Sugon 8000 to complete a 3D high-precision simulation of a 33-nozzle annular array.[ By Zhang Liyan | Tang Ruohan ]