The award will create a new, TOP500-class supercomputer service for AI and high-performance computing (HPC) and enable new research in a wide range of areas, including in clean energy, modelling optimal configuration of wind farms on both land and water, and modelling fusion reactors to provide green energy in the future.
The collaboration with NVIDIA will also mean that Isambard will be able to support cutting-edge research in AI and machine learning and support new user communities from AI, life sciences, medical, astrophysics and biotech.
Isambard 3 will utilise new technologies, including the Arm Neoverse-based NVIDIA Grace CPU Superchip, to provide a production system of at least 55,000 cores. The system - one of the first in the world based on NVIDIA Grace - will have more than six times the computational performance and six times the energy efficiency of Isambard 2.
The new facility will be hosted in a self-cooled, self-contained HPE Performance Optimised Data Center (POD) at the National Composites Centre on the Bristol and Bath Science Park.
Isambard 3 will also include an expanded and upgraded multi architecture comparison system to enable scientifically rigorous performance comparisons and benchmarking across diverse computer architectures and will also feature a storage system comprised of the Cray ClusterStor E1000 storage system to deliver expanded storage with intelligent tiering to support data-intensive workloads, such as AI model training.
Initially hosted by the Met Office to develop more sophisticated weather forecasting and climate prediction modelling, Isambard has also been used to investigate next-generation healthcare and to develop innovations in medicine - researchers are, for example, running molecular level simulations to understand the mechanisms behind Parkinson’s disease, and to help develop new drugs to treat osteoporosis.
Since its start in 2017, Isambard has enabled nearly a thousand people to be trained in advanced computing and simulation techniques all over the world. Isambard 3’s expanded capabilities will create the opportunity for new training around advanced AI workloads through tutorials and hackathons.
Employing the latest sustainability techniques and best practice, Isambard 3 is being designed to be one of the most energy-efficient, lowest carbon emission CPU-based supercomputers in the world, with the potential to reuse waste energy to heat surrounding buildings.
Professor Simon McIntosh-Smith, Principal Investigator for the GW4 Isambard project and a Professor of high-performance computing at the University of Bristol, said, “Isambard is already pushing the boundaries of scientific research, with significant developments being made that just wouldn’t be possible without it. Isambard 3 will create several ‘world-firsts,’ and the new Arm-based NVIDIA Grace processors will enable new kinds of research and innovation in artificial intelligence and scientific simulations.
“Isambard 3 will provide researchers across the UK and internationally access to cutting-edge technology, with a transformational increase in performance and energy efficiency. It’s a great vehicle for supporting collaborations with our academic and industrial partners all over the world.”
John Josephakis, global vice president of business development for HPC and supercomputing at NVIDIA, added, “Isambard 3, powered by the Arm-based NVIDIA Grace CPU Superchip, will deliver unprecedented performance, enabling researchers and scientists to make ground-breaking advances in science and technology.”
It is expected Isambard 3 will be installed later this year, with user migration taking place over winter, ready for use early 2024.
