英伟达将参与开发日本超算“富岳”的后续机型

Core Insights - Nvidia will collaborate with RIKEN to develop AI-oriented GPUs for Japan's next-generation supercomputer, marking the first involvement of a foreign company in Japan's core supercomputer development [1][3] - The new model, codenamed "Fugaku NEXT," aims to achieve the highest global performance in supercomputing by integrating advanced technologies from both Japan and the U.S. [1][3] Development Details - Fujitsu is confirmed to participate in the development, responsible for the overall system design and the central processing unit (CPU), with plans to begin operations around 2030 [3][6] - The project aims to establish Japan's position as a leading AI nation by designing the highest-level CPU and GPU for the AI era [3][6] Performance Goals - The Fugaku NEXT aims for a peak performance of 10 zetta operations per second, significantly enhancing capabilities in AI and advanced simulation computing [5][6] - The new system is expected to improve computational power by 5 to 10 times compared to the existing Fugaku supercomputer [5] Competitive Landscape - The global trend of utilizing AI to accelerate scientific research is intensifying, with the U.S. and China rapidly developing next-generation high-performance supercomputers [9] - Japan aims to enhance its AI capabilities to secure a top-tier position in the international supercomputing arena [9]