May 30, 2022 \ AI, Cloud, HPC
Graphcore to accelerate fusion energy research at Hartree Centre
May 30, 2022 \ AI, Cloud, HPC
The Hartree Centre, a world-leading supercomputing facility dedicated to industry engagement, will use Graphcore’s Intelligence Processing Unit (IPU) technology for its research to deliver clean and commercially viable fusion energy in collaboration with the UK Atomic Energy Authority (UKAEA).
The adoption of IPU systems within the Hartree Centre’s supercomputing cluster marks the start of a strategic partnership between Graphcore and the centre’s parent organisation, the Science and Technology Facilities Council (STFC). The new partnership covers a broader remit related to industrial engagement and development of the UK’s exascale computing competitiveness.
The Hartree Centre’s mission is to transform UK industry through high-performance computing (HPC), data analytics, and AI. Last year the Hartree Centre became the site of a new centre of excellence in extreme scale computing for fusion energy under a collaboration between STFC and UKAEA.
As an international leader in fusion energy science and technology, UKAEA believes fusion will be a safe, low carbon and sustainable part of the world’s future energy supply.
Under STFC’s collaboration with UKAEA, the Hartree Centre is applying the latest computing systems and top-tier supercomputing and data science expertise to run complex models and simulations that will help scientists and engineers progress understanding of fusion and develop viable powerplant technologies. Research projects include developing “digital twins” of future fusion powerplants, and understanding and modelling plasma, the complex, fourth state of matter that drives the fusion process.
Given the prohibitive technical and cost barriers to real-world prototyping, projects like these must be conducted virtually “in-silico,” requiring access to some of the world’s most advanced and capable computing resources.
“The goal of delivering clean fusion energy is one of the great scientific endeavours of our time. Breakthrough technologies in every field are being deployed in pursuit of this goal – including the use of artificial intelligence for simulation,” said Prof. Kate Royse, Director of the Hartree Centre.
“We look forward to working with Graphcore in such an exciting field where accelerating progress could mean that this transformative new way of generating energy is available to everyone, sooner rather than later.”
The STFC-UKAEA fusion energy research projects exemplify the ever-growing trend of convergence between AI and HPC. While HPC is an essential tool fuelling the advance of computational science, AI is increasingly being used alongside more traditional HPC techniques to run large-scale simulations in an accelerated timeframe. Graphcore systems are already being used on data from the Large Hadron Collider.
This convergence between HPC and AI is generating new types of heterogeneous workloads that cannot be satisfied by single monolithic systems; such workloads require new hybrid systems and approaches that combine benefits from both worlds to accelerate and achieve sustained exascale performance.
Graphcore’s IPU systems are particularly well-suited to such hybrid AI-HPC workloads and can deliver class-leading performance over a range of applications while seamlessly fitting into existing datacentres.
“Modelling the hugely complex, strongly coupled, multi-physics system that is a fusion powerplant is undeniably a simulation grand challenge – an endeavour that has long been heralded an 'exascale' challenge,” said Rob Akers, Head of Advanced Computing at UKAEA.
“Only now, as the world’s first exascale machines become available, and through the convergence of AI methods and traditional HPC algorithms is digitally twinning of fusion powerplant technology becoming a realistic prospect. Graphcore’s IPU systems are a tantalising glimpse of what a second generation of exascale machines might look like. We are incredibly excited to be working with both STFC Hartree Centre and Graphcore to engineer a transformation around our predictive capability for fusion powerplant design.”
For the first group of Hartree Centre researchers working with Graphcore, access to IPU technology will be provided via the IPU Cloud service offered by G-Core Labs, a European Cloud and EDGE provider. Atos, a global leader in AI and HPC technology, will work closely with G-Core Labs, Graphcore, and the Hartree Centre’s researcher team as a systems integrator.
As part of the strategic partnership between Graphcore and STFC, Graphcore will provide extensive technology training (including workshops), while STFC will share guidance and feedback related to the performance of IPU systems in speeding up workloads and accelerating research at the Hartree Centre.
STFC, Graphcore, and Atos will also explore further partnerships to support the UK’s exascale computing strategy, which aims to provide appropriate and ambitious compute capabilities for diverse research and innovation communities.
“Atos is delighted to be working with Graphcore and G-Core Labs to help provide STFC with advanced AI compute to support their vital work on nuclear fusion energy, and beyond. This sort of essential, world-changing research exemplifies the best of what converged AI and HPC can enable,” said Andy Grant, VP, Global Head of HPC, AI and Quantum Sales, Atos.
Additionally, the partnership calls for STFC, Graphcore and Atos to jointly explore industrial partnerships and engagements—including though the newly-established Hartree National Centre for Digital Innovation (HNCDI), which provides a safe and supportive environment for businesses to acquire the skills, knowledge and technical capability required to adopt digital technologies like supercomputing, data analytics, AI, and quantum computing.
Sign up for Graphcore updates:
Mar 24, 2023 \ NLP, Large models, PaperspaceRead Article
Mar 24, 2023 \ Developer, NLP, Large models, PaperspaceRead Article
Feb 16, 2023 \ Drug Discovery, Scientific Research, GNNRead Article
Sign up below to get the latest news and updates: