The Japanese research institute RIKEN has adopted NVIDIA's latest technology for scientific artificial intelligence and Quantum Computing.

NVIDIA announced that Japan's leading national research institute RIKEN has integrated the NVIDIA GB200 NVL4 system into its new supercomputer, used for scientific artificial intelligence and Quantum Computing, accelerating research in Japan in the fields of life sciences, materials science, climate and weather forecasting, manufacturing, laboratory automation, and quantum applications, marking a significant milestone in the upgrade of Japan's scientific research infrastructure.

1600 Blackwell GPUs accelerate scientific research

The first new system in the RIKEN “Scientific Artificial Intelligence” project will be equipped with 1,600 NVIDIA Blackwell GPUs, built on the GB200 NVL4 platform, connected via NVIDIA Quantum-X800 InfiniBand. This system aims to accelerate multi-disciplinary research, including precision medicine, materials innovation, climate simulation, industrial manufacturing optimization, and scientific research automation, to support Japan in establishing a more autonomous scientific AI capability.

Quantum Computing platform takes shape, driven by 540 GPUs for hybrid quantum simulation.

The second supercomputer from RIKEN will be used for quantum computing research, featuring 540 Blackwell GPUs, and also utilizing GB200 NVL4 and Quantum-X800 networking technologies. This platform will enhance workloads such as quantum algorithms and quantum-classical hybrid simulations, assisting research teams in accelerating breakthroughs in quantum computing applications.

NVIDIA collaborates with RIKEN to build a new generation of computing power infrastructure.

Satoshi Matsuoka, director of the RIKEN Center for Computational Science, stated that the integration of NVIDIA's GB200 NVL4 will lead to a world-leading unified platform for AI, Quantum Computing, and HPC, providing unprecedented momentum for scientific research and various social applications.

Supercomputing Blueprint Expansion: FugakuNEXT and Two New GPU Systems Development

After Fujitsu and NVIDIA announced their collaboration to create FugakuNEXT in August, RIKEN further activated these two GPU accelerated systems as a proxy platform for developing the next generation flagship supercomputer. The FugakuNEXT project will use the FUJITSU-MONAKA-X CPU and achieve high-bandwidth interconnection with NVIDIA GPUs through NVIDIA NVLink Fusion technology. This future supercomputer is expected to reach application performance that is one hundred times that of existing CPU systems and will have the capability to integrate Quantum Computing.

Japan's scientific research capabilities are comprehensively upgraded, with the convergence of AI, HPC, and Quantum Computing.

Combining the architecture design of MONAKA-X and NVIDIA's latest GPUs, FugakuNEXT will become the core infrastructure driving Japan's continued leadership in high-performance computing, artificial intelligence, and Quantum Computing. The increasingly close cooperation between RIKEN and NVIDIA also symbolizes Japan's determination to strengthen its independent innovation capabilities in global technological competition.

CUDA-X and GPU-Accelerated Scientific Simulation

In addition to hardware upgrades, NVIDIA and RIKEN are also collaborating on the development of floating-point simulation technology accelerated by Tensor Core, enabling traditional scientific computing to effectively leverage the advantages of GPUs. RIKEN plans to further utilize the CUDA-X suite (which includes over 400 GPU-accelerated libraries and tools) to enhance the performance of HPC and scientific applications, strengthening Japan's international competitiveness in scientific AI and quantum research.

This article discusses how the Japanese research institute RIKEN has introduced NVIDIA's latest technology for scientific artificial intelligence and Quantum Computing, first appearing on Chain News ABMedia.