Computer Science GPU Cluster
The Computer Science GPU Cluster provides significant, primarily GPU, compute capability hosted within the School itself. It is a Slurm-based cluster: batch jobs can be bundled together and submitted to run for hours, days or weeks without needing to be supervised. It is presented as a solid alternative and/or complement to the University's central HPC solution, Ada.
The CS cluster currently comprises 300 CPU cores and 56 GPUs with up to 48GB VRAM per card (and 2x 96GB cards coming online soon) and 500Gb RAM average per node. It is supported by 300TB storage and operates at 13K TFLOPS. In-house technical support is available, along with clear written guidance.
Our GPU facility developed within the Computer Vision Laboratory, but is now a School resource, supporting both research and teaching.