The Durham AMD GPU testbed provides researchers with the opportunity to test their codes on AMD MI50, MI100 and (from Q4 2022) MI210 GPUs. The testbed consists of separate servers hosting each generation of GPUs each with 1TB RAM and 32 or 64 cores per node.
AMD software, ROCM, AOCC, AOMP and GCC with offload support are installed, other tools can be made available upon request.
The GPU testbed extends the existing AMD cluster at Durham and is already in use by researchers at various sites including the University of Bristol and the Hartree Centre, with the first journal submission for research using this testbed already submitted.
This system is integrated with COSMA allowing the existing login nodes and infrastructure to be used. Users are able to request an account through the SAFE system managed by EPCC, and should contact Alastair Basden to discuss their requirements.