DISS: Computing Resources

For MSc students who need access to GPUs, there are several options.

ICF cluster

ICF stands for the Informatics Computing Facility, and the cluster is maintained by the School of Informatics. This should be the default option if your project needs to use GPUs. This was the old MLP cluster, but now expanded. Most of the GPUs have 12GB of memory (NVIDIA GeForce RTX 2080). While there are some cards that have more memory, they are in high demand. Please discuss with your project supervisor if you think you need GPUs with more than 12 GB of memory. To learn more about using the cluster, see this tutorial.

Eddie

Eddie is a cluster maintained by the Edinburgh Compute and Data Facility (ECDF). There are around 80 GPU cards on Eddie, but the cluster is shared among the entire university. There is also the extra hassle of setting up the storage for using the cluster. Eddie is better suited if your project requires a lot of *CPU* compute. You can learn more about Eddie in the wiki here.

EIDF GPU cluster

The EIDF GPU cluster is maintained by the Edinburgh International Data Facility (EIDF) under the Edinburgh Parallel Computing Centre (EPCC). The EIDF cluster has many high-end GPUs, such as A100, H100, and H200, that are *only* for training large-scale models. The EIDF cluster is not open to taught MSc students, and most MSc projects do not require these high-end GPUs. Please discuss with your supervisor if you think you need Documents about the EIDF cluster can be found on this git repo.