Nvidia’s JUPITER: Europe's Fastest Exascale AI Supercomputer

Nvidia has deployed its Grace Hopper platform to power Europe's fastest supercomputer at the Jülich Supercomputing Centre in Germany.

The JUPITER system delivers more than twice the performance for high-performance computing and AI workloads compared with the next fastest system, according to the company. 

Now among the top five systems on the TOP500 list of the world's fastest supercomputers – JUPITER achieves the highest energy efficiency at 60 gigaflops per watt – a gigaflop being one billion floating-point operations per second, a measure of computational speed.

“AI will supercharge scientific discovery and industrial innovation,” says Jensen Huang, Founder and CEO of Nvidia.

“In partnership with Jülich and Eviden, we're building Europe's most advanced AI supercomputer to enable the leading researchers, industries and institutions to expand human knowledge, accelerate breakthroughs and drive national advancement.”

How JUPITER targets exascale computing performance

The system is owned by the EuroHPC Joint Undertaking, a European Union initiative that coordinates supercomputing efforts across member states. Eviden, part of the Atos Group technology services company, also provided the system architecture.

JUPITER is positioned to become Europe's first exascale supercomputer, meaning it can perform one quintillion calculations per second. Exascale computing is a thousand-fold increase over petascale systems and enables simulations of unprecedented complexity.

The system comprises nearly 24,000 Nvidia GH200 Grace Hopper Superchips connected through the company's Quantum-2 InfiniBand networking platform – InfiniBand being a high-speed networking standard used in supercomputing environments to connect processors and storage systems.

The supercomputer is also expected to reach over 90 exaflops of AI performance. An exaflop measures computational performance in AI workloads, specifically operations optimised for machine learning tasks rather than traditional scientific computing.

“With JUPITER's extreme performance, Europe has taken a giant leap into the future of science, technology and sovereignty,” says Anders Jensen, Executive Director of the EuroHPC Joint Undertaking.

“JUPITER's computing power will serve as a catalyst for scientific discovery, propelling foundational research across the continent in fields as diverse as climate modelling, energy systems and biomedical innovation.”

The system additionally uses Eviden's BullSequana XH3000 liquid-cooled architecture to manage the heat generated by the high-performance processors. 

Nvidia’s Grace Hopper platform powering scientific applications

The Grace Hopper platform combines Nvidia's Grace central processing unit with its Hopper graphics processing unit (GPU) architecture – targeting applications that process large datasets, offering up to 10 times higher performance for workloads handling terabytes of data.

As a result, JUPITER enables faster simulation, training and inference of large AI models across multiple scientific disciplines. 

Furthermore, the system supports climate and weather modelling through Nvidia's Earth-2 platform, which creates high-resolution environmental simulations – contributing to the Earth Virtualisation Engines project, which aims to create a digital twin of Earth's climate system.

For quantum computing research, JUPITER provides access to Nvidia's CUDA-Q platform and cuQuantum software development kit.

“JUPITER is a landmark achievement for European science and technology,” says Thomas Lippert, co-director of the Jülich Supercomputing Centre.

“Powered by Nvidia's accelerated computing and AI platforms, JUPITER is advancing the frontier of foundation model training and high-performance simulation, enabling researchers across Europe to tackle challenges of unprecedented complexity.”

The supercomputer also supports computer-aided engineering through Nvidia's PhysicsNeMo framework, CUDA-X libraries and Omniverse platform. These tools enable AI-driven simulation and digital twin technologies for product design and manufacturing applications.

JUPITER’s abilities for pharmaceutical research 

In drug discovery, JUPITER utilises Nvidia's BioNeMo platform to accelerate the creation and deployment of AI models for pharmaceutical research.

This way, the platform streamlines biomolecular science applications, reducing the time required to generate insights from complex biological data.

The system is what Nvidia characterises as an AI Factory, a data centre designed specifically for AI workloads. 

“JUPITER will substantially advance quantum algorithms and hardware development,” says Kristel Michielsen, Co-Director of the Jülich Supercomputing Centre.

“Hybrid quantum HPC-computation will profit from powerful tools such as the Nvidia CUDA-Q platform and the Nvidia cuQuantum software development kit.”

More broadly, the deployment is a collaboration between multiple European organisations and demonstrates the region's investment in maintaining competitiveness in high-performance computing and AI research.

Emmanuel Le Roux, Senior Vice President and Global Head of Advanced Computing at Eviden, Atos Group, says: “JUPITER's launch is not just an extraordinary technical success – delivering an exascale machine and Julich's modular data centre in less than nine months – it marks a pivotal moment for European high-performance computing.

“It clearly demonstrated the technological leadership of the European Eviden-led consortium, which designed, built and delivered this world-class system.”

Explore the latest edition of AI Magazine and be part of the conversation at our global conference series, Tech & AI LIVE. Discover all our upcoming events and secure your tickets today.

Also sign up to our free weekly newsletter for the latest insights and stories straight into your inbox.

AI Magazine is a BizClik brand