The University of Manchester is one of the N8 sites contributing to maintain the full service on the N8 CIR Bede system for another year. The current funding came to an end in July 2026 but 6 of the 8 sites have come together to maintain the service for their researchers.

Now that the previous funding has come to an end, there is no longer an allocation for external users, so Manchester users will now have an increased fair share - a sixth of the resources instead of an eighth as previously.

While the Grace-Hopper nodes are over-subscribed the V100 nodes are currently under-subscribed — i.e. provide good throughput, probably compared with the CSF GPUs. Thus we encourage you to apply to use the service if you think you can make use of the facility to help spread the load. Particularly as the largest site, we would like to see Manchester using its fair share of available resources!

The V100s are still capable, particularly for double precision, with relatively high bandwidth to the CPUs and unified memory. The T4 nodes are specialized for inference, but may be useful for other work; Gromacs has run acceptably on them.

The Infiniband network is suitable for multi-node work if you have a scalable application, or want to test potential scalability to systems like the national AIRR facilities. Anything that used to run on the previous-generation Summit and Sierra supercomputers should work well since the architecture is the same — n.b. not x86_64.

Limited support is available from our Research Platforms team to get your research up and running as well as the N8 CIR support desk.

You can also join in the discussion in our GPU User Group - join the group's Team space to share your thoughts.