UK details Isambard AI supercomputer
The UK’s AI supercomputer will have over 5000 Grace Hopper processors from Nvidia, which will provide up to 21 exaFLops of performance and be one of the most powerful systems in Europe.
Isambard AI is planned for the summer of 2024 and will be built by Hewlett Packard Enterprise (HPE) who has built previous systems. The HPE Cray EX supercomputers will have over 5,000 Nvidia GH200 superchips.
The facility at the National Composite Centre in Bristol will be a self-cooled, self-contained data centre and will be used for training large language models (LLMs), big data and robotics as well as by the UK’s new AI Task Force.
- Bristol to host UK’s exascale AI supercomputer
- SiPearl wins first deal for ARM exascale supercomputer
- Swiss AI supercomputer to use new Nvidia ARM chip
Isambard-AI will connect with a new supercomputer cluster at the University of Cambridge, called Dawn, which is being developed to offer additional capacity as part of the new national AIRR.
The GH200 superchip combines the 72-core Grace CPU with 480 GBytes of ECC LPDDR5X memory and a GH100 computing GPU with 141 GBytes of HBM3E memory.
Other AI supercomputer systems being developed include Inflection AI which is also using the H100 to build a cluster with 3500 chips for 22exaFlops of AI performance. Tachyum in Slovakia is also building an AI supercomputer in the US with its Prodigy chip that can reach 50 exaFLops.
The Condor Galaxy 1 (CG-1) from waferscale chip designer Cerebras is a 64-node AI supercomputer with 54 million cores and 82Tbytes of data that can deliver 4 exaFLOPs of processing being built in the United Arab Emirates. The roadmap sees the rollout of performance up to 36 exaFLOPS,
- Cerebras delivers 4exaFLOP AI supercomputer
- Nvidia, Microsoft team on massive cloud AI supercomputer
“Isambard-AI represents a huge leap forward for AI computational power in the UK. Today Isambard-AI would rank within the top 10 fastest supercomputers in the world and, when in operation later in 2024, it will be one of the most powerful AI systems for open science anywhere,” said Professor Simon McIntosh-Smith, Director of the Isambard National Research Facility at the University of Bristol.
“It’s immensely exciting to be at the forefront of the AI revolution and to partner with industry leaders HPE and Nvidia to rapidly build and deploy large-scale research computing infrastructure to create one of the most powerful supercomputers in the world. Isambard-AI will offer capacity never seen before in the UK for researchers and industry to harness the huge potential of AI in fields such as robotics, big data, climate research and drug discovery.”
Justin Hotard, Executive Vice President and General Manager, HPC, AI & Labs at HPE, said: “Today’s announcement of the UK’s major investment in AI supercomputing underscores its commitment to taking a global leadership position in AI. The Isambard-AI system will harness world-leading supercomputing, including high-performance networking co-developed at HPE’s Bristol labs, to provide the performance and scale required for compute-intensive AI projects. We are proud to partner with the UK Government and the University of Bristol to give UK researchers and industry access to Europe’s largest AI system for open science.”
Ian Buck, Vice President of Hyperscale and HPC at Nvidia, said: “In building one of the world’s fastest AI supercomputers, the UK is demonstrating the importance for nations to create their own infrastructure. Isambard-AI will provide researchers with the same state-of-the-art AI and HPC compute resources used by the world’s leading AI pioneers, enabling the UK to introduce the next wave of AI and scientific breakthroughs.”
The government’s new Frontier AI Taskforce, also announced today at the AI Safety summit, will have priority access to support its work to mitigate the risks posed by the most advanced forms of AI, including national security from the development of bioweapons and cyberattacks. The resource will also support the work of the AI Safety Institute, as it develops a programme of research looking at the safety of frontier AI models and supports government policy.
Phase one of the system, available in March 2024, will use Isambard 3, which is due to be installed at the NCC at the start of 2024 and was developed in collaboration with the GW4 group of universities – an alliance made up of the Universities of Bath, Bristol, Cardiff and Exeter. It will offer early access to UK-based scientists, researchers and developers so their research can get underway as soon as possible.
The University of Bristol has been awarded funding to invest in a new Centre for Doctoral Training (CDT) in Artificial Intelligence on the back of the Isambard AI supercomputer.
The CDT is one of 12 centres to receive a share of a £117m from the UK Research and Innovation (UKRI) and aims to ensure the UK has the skills needed for AI.
PhD students who train at the CDT, called the UKRI AI Centre for Doctoral Training in Practice-Oriented Artificial Intelligence (PrO-AI), will learn how to design and manage the entire lifecycle of advanced AI applications in science and research, developing AI solutions for scientific problems in a safe and transparent manner.
The centre complements the existing UKRI Centre for Doctoral Training in Interactive Artificial Intelligence, which was launched in 2019 and builds on will build on Bristol’s unique strengths in intelligent systems, machine learning, and human-computer interaction.
“Putting applications centre stage is a natural and necessary next step now that AI technology is becoming increasingly mature,” said Professor Peter Flach, Professor of Artificial Intelligence at the University of Bristol. “The award of this CDT and the recently announced Isambard-AI supercomputer confirms Bristol’s reputation as an internationaentre of excellence in cutting-edge AI research.”
The first cohort of UKRI AI CDT students will start in the 2024/2025 academic year and recruitment for which will begin shortly.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
