Research Computing Network Launch: AI & GPU Frontiers

About the Event

Join us on Tuesday, 17th June 2025 for the Research Computing Network Launch: AI & GPU Frontiers, a full-day internal event designed to strengthen the University of Leeds research computing community — with a focus on GPU-driven research.

This is a strategic academic gathering where we aim to share, shape, and align our computing infrastructure with real research needs. We’ll explore current uses of AI, machine learning, scientific computing, and GPU-accelerated workloads across disciplines at Leeds. We also want to hear your future needs and expectations to better guide our infrastructure planning.

Event Details

🗓 Date: Tuesday, 17th June 2025
🕤 Time: 9:30 AM – 4:00 PM (includes lunch and networking)
📍 Location: Liberty Building, Belle Vue Road, University of Leeds, LS2 9JT
🎟 Registration: Free (internal attendees only)

Shaping the Future of Research Computing Together

This is your opportunity to shape what comes next. We’re inviting you — our academic colleagues — to tell us:

• What are you doing with GPU-based computing?
• What are your main takeaways?
• What’s stopping you from going further?
• What do you wish our systems could do?

Your answers will directly influence the design of our next HPC system (Calder) and the one after that (Derwent). Our goal is to build systems that are tailored to research needs.

What You’ll Get from This Event

• Hear how Leeds is investing in GPU infrastructure to meet research needs
• See what peers across the university are doing with AI, simulation, and GPU-based workflows
• Meet and talk with Nvidia and Dell representatives
• Contribute your voice to future hardware and service design
• Share your work, challenges, or even your "wish list" for future systems
• Build connections with others facing similar research and computing questions

What We’re Asking From You

We invite short talks and posters from researchers working with, or planning to use, GPU-powered computing. You can:

• Present ongoing or past work using GPUs (e.g. for AI, deep learning, simulations)
• Share planned research that needs future capability
• Tell us what’s missing in current infrastructure
• Offer feedback on how systems and services could better support you

Don’t worry if your project isn’t finished — ideas and roadblocks are just as valuable.

📅 Deadline: 1st June 2025
📬 Notification: 3rd June 2025

Why Now?

This event is part of the Research IT (RIT) Project, a 5-year university-wide effort to transform digital research infrastructure. RIT’s mission is to position our University at the heart of global research and innovation by delivering world-class, sustainable, and responsive research technology platforms.

The RIT roadmap covers Support, HPC, Research Storage, General Computing, and Sustainable Funding. The main focus of this event is our HPC strategy, especially around GPU provision.

Our GPU Journey So Far

Leeds has historically relied on ARC3 and ARC4 HPC systems, providing limited GPU capability (40 GPUs total across both). In early 2025, we launched Aire, a £2.4M HPC system with 84 × NVIDIA L40S GPUs. While powerful and well-received, L40S chips aren’t ideal for all workloads, particularly those requiring high-precision (FP64) performance.

Aire was born out of urgent need and a heated GPU market — and it was a success. But it also highlighted the importance of planning ahead based on academic needs, not hardware constraints. That’s why we’re moving to a new model: rather than launching a new HPC every 5 years, our strategy is to release a new system every 12–18 months. This will:

• Provide continuous access to cutting-edge hardware
• Minimise risks related to hardware failures
• Better match emerging research demands
• Enable smarter, future-proof infrastructure decisions

Looking Ahead: Calder and Derwent

Our next system, Calder, is now in tender, with a value of £3.6M, designed to complement Aire’s capabilities. Calder will balance CPU and GPU investment, with a strong preference for FP64-capable GPUs (e.g., NVIDIA H or B series), supporting a broader range of workflows.

As soon as Calder is awarded (summer 2025), we’ll begin work on its successor, Derwent, to stay on track with our rolling upgrade strategy.

Contact

This meeting is organised by Leeds’s Research IT in collaboration with the Library and SciML Community. Enquiries can be sent to rcteam@leeds.ac.uk.

Conference Committee

• Patricia Ternes – Research IT, University of Leeds
• Zoe Hancox - Leeds Institute of Health Sciences, University of Leeds
• Marc de Kamps - School of Computer Science, University of Leeds
• Claire Knowles - Library, University of Leeds

Supporters

This event is supported by

• NVIDIA
• Dell Technologies

Agenda