Japan is as soon as once more constructing a landmark high-performance computing system — not just by chasing pace, however by rethinking how expertise can finest serve the nation’s most pressing scientific wants.
On the FugakuNEXT Worldwide Initiative Launch Ceremony held in Tokyo on Aug. 22, leaders from RIKEN, Japan’s prime analysis institute, introduced the beginning of a world collaboration with Fujitsu and NVIDIA to co-design FugakuNEXT, the successor to the world-renowned supercomputer, Fugaku.
Awarded early within the course of, the contract permits the companions to work aspect by aspect in shaping the system’s structure to deal with Japan’s most crucial analysis priorities — from earth techniques modeling and catastrophe resilience to drug discovery and superior manufacturing.
Greater than an improve, the hassle will spotlight Japan’s embrace of contemporary AI and showcase Japanese improvements that may be harnessed by researchers and enterprises throughout the globe.
The ceremony featured remarks from the initiative’s leaders, RIKEN President Makoto Gonokami and Satoshi Matsuoka, director of the RIKEN Middle for Computational Science and one in all Japan’s most revered high-performance computing architects.
Fujitsu Chief Expertise Officer Vivek Mahajan attended, emphasizing the corporate’s position in advancing Japan’s computing capabilities.
Ian Buck, vice chairman of hyperscale and high-performance computing at NVIDIA, attended in particular person as nicely to debate the collaborative design method and the way the ensuing platform will function a basis for innovation nicely into the subsequent decade.
Momentum has been constructing. When NVIDIA founder and CEO Jensen Huang touched down in Tokyo final yr, he known as on Japan to grab the second — to place NVIDIA’s newest applied sciences to work constructing its personal AI, by itself soil, with its personal infrastructure.
FugakuNEXT solutions that decision, drawing on NVIDIA’s entire software program stack — from NVIDIA CUDA-X libraries resembling NVIDIA cuQuantum for quantum simulation, RAPIDS for information science, NVIDIA TensorRT for high-performance inference and NVIDIA NeMo for big language mannequin growth, to different domain-specific software program growth kits tailor-made for science and business.
Improvements pioneered on FugakuNEXT may turn out to be blueprints for the world.
What’s Inside
FugakuNEXT will probably be a hybrid AI-HPC system, combining simulation and AI workloads.
It can function FUJITSU-MONAKA-X CPUs, which may be paired with NVIDIA applied sciences utilizing NVLink Fusion, new silicon enabling high-bandwidth connections between Fujitsu’s CPUs and NVIDIA’s structure.
The system will probably be constructed for pace, scale and effectivity.
What It Will Do
FugakuNEXT will assist a variety of functions — resembling automating speculation era, code creation and experiment simulation.
- Scientific analysis: Accelerating simulations with surrogate fashions and physics-informed neural networks.
- Manufacturing: Utilizing AI to be taught from simulations to generate environment friendly and aesthetically pleasing designs quicker than ever earlier than.
- Earth techniques modeling: aiding catastrophe preparedness and prediction for earthquakes and extreme climate, and extra.
RIKEN, Fujitsu and NVIDIA will collaborate on software program developments, together with instruments for mixed-precision computing, steady benchmarking, and efficiency optimization.
FugakuNEXT isn’t only a technical improve — it’s a strategic funding in Japan’s future.
Backed by Japan’s MEXT (Ministry of Training, Tradition, Sports activities, Science and Expertise), it’ll serve universities, authorities companies, and business companions nationwide.
It marks the beginning of a brand new period in Japanese supercomputing — one constructed on sovereign infrastructure, international collaboration, and a dedication to scientific management.
Picture courtesy of RIKEN