Accelerating research computing at Oxford

High-performance computing (HPC) is the backbone of advanced university research, enabling breakthroughs across disciplines such as computational biology, artificial intelligence, and climate modelling.  

With recent enhancements bringing powerful general-purpose graphical processing units (GPUs) and new storage for high-performance scratch space, Oxford’s Advanced Research Computing (ARC) team is providing a cutting-edge platform for researchers, optimising both computational power and data access. 

Advanced computing for diverse workloads 

Coming very soon, two new high performance GPU nodes will provide a platform for tackling the most demanding computational challenges, enabling researchers to seamlessly run workloads ranging from large-scale simulations to advanced AI models, all within the same infrastructure. 

Work is ongoing to deliver these new GPUs which will deliver unparalleled performance. Each new GPU will offer up to 3 teraflops of FP64 computing power, excelling in AI inference and providing up to 700 teraflops of performance (approximately the combined power of around 7000 smart TVs or 140 laptops).

High-performance scratch space for data-intensive workloads 

To complement the computational power of our new GPUs, we’ve recently deployed high performance scalable storage as the primary scratch space for our HPC cluster. Scratch space is crucial for temporary data storage during computation, where performance and speed are essential. The new platform delivers up to 80 GB/s of throughput per client and ultra-low latency, ensuring data can be accessed and processed without bottlenecks. 

Integrated with our current workload manager, the new scratch storage dynamically supports job scheduling, allowing seamless data access during computations. This ensures researchers can run simulations, analyse large datasets, or train machine learning models without interruptions. 

More information and how to access this service 

For more information about high performance computing at Oxford and how to sign up, visit www.arc.ox.ac.uk or email your questions to support@arc.ox.ac.uk