AIRR Expression of Interest to use UK large scale compute
Introducton
UK Research and Innovation (UKRI), on behalf of the Department for Science, Innovation and Technology (DSIT), invites researchers and innovators from across the UK to express their interest in accessing large-scale AI compute, in particular, the new AI Research Resource, the Isambard-AI and Dawn compute services.
We are seeking Expressions of Interest (EoIs) from researchers and innovators who can demonstrate a clear need for AI compute and may be suitable for early access to the Isambard AI and Dawn services as part of their testing phase.
About The AI Research Resource (AIRR)
The AIRR comprises Isambard-AI and Dawn, AI-optimised HPC platforms designed to support ambitious research projects. They provide access to advanced hardware, software, and expert support to enable researchers to tackle complex challenges and achieve groundbreaking results. The AIRR is intended to provide a step change in the compute available to UK researchers. As such, we would like researchers to start considering large-scale projects that can be run on these systems.
The Expression of Interest (EoI) phase
The AIRR systems will come online in early 2025 and are currently in the early stages of being tested and installed. Access to the AIRR compute will be opened using a phased approach, the first being this EoI, which will provide opportunities to test the system up until its launch. This will enable us to monitor user demand and inform the development of services alongside the commissioning of new systems. The objectives of this initial EoI phase are to:
- identify parties interested in using AIRR
- identify potential pilot projects suitable for early access to Isambard AI and Dawn during the commissioning period
- understand user demand
- help design access process to suit a range of users, and
- understand the range of different user requirements.
These insights will be used to design the first main call for AIRR users, which will launch early in 2025.
Responding to this EoI does not commit you to any further activities, and any decision not to engage at this stage will not prevent you from applying once the full application process for large-scale AI compute is launched. Applicants should be aware that the AIRR systems are still undergoing testing. As such they are not yet prepared to offer the breadth of user support that users might expect from a national compute service.
The EoI form opens on 14 January 2025. The EoI process will not have a fixed end date. Applications will be batched and assessed on a regular basis.
Who should fill this in
This EoI is open to all UK-based researchers and innovators, however, we are particularly keen to hear from projects that contribute to delivering against the government’s five missions: growing the economy, an NHS fit for the future, safer streets, opportunity for all, and making Britain a clean energy superpower.
Some potential projects focused on these might include:
- Applying foundation models to particular scientific disciplines, such as climate science or biology
- Developing autonomous scientists capable of generating novel hypotheses, coding up basic experiments, analysing results, and increasing research productivity
- Exploring new paradigms in basic AI research such as in AI safety or novel architectures.
We are also accepting applications from other fields including mixed workloads incorporating other techniques.
This EoI consists of three pages:
- Your contact details (this page, required)
- Details of your interest in using large-scale AI compute (required)
- Details of your potential pilot project, if applicable (optional).
We are interested in hearing about pilot projects using the Isambard AI and Dawn services from researchers from UK universities, research institutions, and businesses.
More Information
Isambard-AI has 40 compute nodes, each of which contains 4 Nvidia Grace Hopper (GH200) superchips. Each node has 288 Grace CPU cores and 4 H100 GPUs. There is 512 GB of CPU memory per node, and 384 GB of High Bandwidth (GPU) memory. The nodes are connected using a Slingshot high performance network interconnect (4 200 Gbps injection points per node).
From summer 2025, users will also be able to access Isambard-AI phase 2 through the early access call while the system is being tested, which has an additional 5,280 Nvidia Grace Hopper (GH200) superchips.
More Information
Dawn consists of 256 Dell XE9640 server nodes, each server has 4 Intel Data Centre Max 1550 GPUs (each GPU has 128 GB HBM RAM configured in a 4-way SMP mode with XE-LINK). In total there are 1024 Intel GPUs. Each server has 2 Gen 5 XEONs, and 1TB RAM and 4 HDR200 Infiniband links connected to a fully non-blocking fat tree. There is 14TB of local NVMe storage on each server. Dawn also has 2.8PB of NVMe flash storage. This consists of 18 quad-connected HDR infiniband servers providing 1.8 TB/s of network bandwidth to 432 NVMe drives designed to match the network performance. This High Performance storage layer will be tightly integrated with the scheduling software to support complex pipelines and AI workloads. Dawn has also access to 5PB of HPC Lustre storage on spinning disks. During the pilot phase 100 nodes will be available to users whilst development and performance work continues on the rest of the cluster.
More Information
We are processing, including collecting, your personal data so we can:
- investgate the potential interest in, and possible range of applications to, a future UKRI opportunity for AI compute services
- identify potential early access pilot projects for the Isambard-AI and Dawn services.
We will store your data securely and handle it in line with UK data protection legislation.
When you take part in this EoI process UKRI will process any personal details you chose to share during the course of the survey. The personal information that you provide as part of this survey will only be used in the analysis of the EoI submissions, assessment of any pilot projects and will not be used/shared for any other purpose.
Individual responses will only be viewed by UKRI and DSIT staff. However, UKRI reserves the right to publish and share aggregated information with stakeholders.
Potential early access pilot project details will be shared with the AIRR sites for technical feasibility and with a DSIT assessment panel to ensure a range of topics (including links to government missions and research quality) are covered in the projects assigned early access time.
If you would like to know more, including how to exercise your data protection rights, please see our privacy notice.