RIKEN, Fujitsu, and NVIDIA have launched the development of “Fugaku NEXT,” the successor to Japan’s flagship supercomputer Fugaku. Scheduled to begin operation around 2030, Fugaku NEXT will be the first Japanese national supercomputer to adopt GPUs for acceleration, with NVIDIA designing the GPU infrastructure and Fujitsu leading system and CPU development.
The project aims to create an “AI-HPC platform.” This platform will merge top-tier simulation performance with advanced AI capabilities. It will allow both to work together seamlessly. The new system will build on Fugaku’s CPU technology and add GPU acceleration. This will enhance efficiency in scientific simulations and AI-driven research. 理化学研究所’s software expertise, 富士通’s hardware design, and エヌビディア’s GPU technologies will support this international effort.
こちらもお読みください: SanDisk and SK hynix Team Up on HBF Memory Standard
Fugaku NEXT is expected to achieve up to 100 times the application performance of Fugaku. This will boost advancements in areas like climate science and drug discovery. The initiative will also look into integrating quantum computing, keeping pace with global trends in hybrid computing. The basic design should finish by fiscal year 2025, followed by detailed design work in 2026.