RIKEN Supercomputer: Fujitsu & NVIDIA Partnership
- Japan is embarking on a new era of high-performance computing with FugakuNEXT, a project designed not just for speed, but for addressing critical national research priorities.
- FugakuNEXT is the planned successor to Fugaku, a supercomputer that previously held the title of the world's fastest.
- FugakuNEXT isn't about achieving a raw speed benchmark; it's about maximizing scientific output.
“`html
FugakuNEXT: Japan Launches Enterprising Supercomputer Project with NVIDIA and Fujitsu
Table of Contents
Japan is embarking on a new era of high-performance computing with FugakuNEXT, a project designed not just for speed, but for addressing critical national research priorities. This international collaboration between RIKEN, Fujitsu, and NVIDIA signals a strategic shift towards AI-driven scientific finding.
What is FugakuNEXT?
FugakuNEXT is the planned successor to Fugaku, a supercomputer that previously held the title of the world’s fastest. Announced on August 22, 2023, at the FugakuNEXT International Initiative launch Ceremony in Tokyo, the project represents a significant investment in Japan’s scientific future. Unlike a simple hardware upgrade, FugakuNEXT will be co-designed with a focus on specific research challenges.
Key Objectives and Research areas
FugakuNEXT isn’t about achieving a raw speed benchmark; it’s about maximizing scientific output. The project will prioritize research in four key areas:
- Earth Systems Modeling: Improving climate predictions and understanding complex environmental changes.
- Disaster Resilience: Developing tools for predicting and mitigating the impact of natural disasters, crucial for a nation prone to earthquakes and typhoons.
- Drug Discovery: Accelerating the identification and development of new pharmaceuticals.
- advanced Manufacturing: Optimizing manufacturing processes and creating new materials.
These areas align with Japan’s national priorities and reflect a commitment to using technology for societal benefit. The focus on these specific challenges will drive the system’s architecture and software development.
The Collaboration: RIKEN, Fujitsu, and NVIDIA
The success of FugakuNEXT hinges on the collaborative expertise of its partners.
- RIKEN: Japan’s largest comprehensive research institution, providing the scientific vision and use cases. RIKEN’s Center for Computational Science, led by Satoshi Matsuoka, will play a central role in defining the system’s requirements.
- Fujitsu: A leading Japanese IT company, responsible for system integration and hardware development. Fujitsu’s experience with Fugaku provides a strong foundation for this next-generation project.
- NVIDIA: A global leader in AI and accelerated computing, contributing its expertise in GPU technology and AI software.NVIDIA’s involvement signifies the importance of AI in the future of supercomputing.
This partnership represents a shift towards co-design, where hardware and software are developed in tandem to optimize performance for specific workloads. Ian Buck, NVIDIA’s VP of hyperscale and HPC, emphasized the collaborative design approach.
Technical Details and Expected Performance
While specific performance targets haven’t been released,FugakuNEXT is expected to significantly surpass its predecessor in AI-related workloads. Fugaku achieved 442.01 petaflops on the LINPACK benchmark,but FugakuNEXT will prioritize performance on AI models and data analytics.The system will leverage NVIDIA’s latest GPU technology and fujitsu’s expertise in system architecture.
The architecture will likely incorporate a heterogeneous computing model,combining CPUs and GPUs to accelerate different types of calculations. This approach is becoming increasingly common in modern supercomputers, as it allows for greater versatility and efficiency.
| Metric | Fugaku | FugakuNEXT (Projected) |
|---|---|---|
| LINPACK
|