BlueGem has been configured with three queues;

• cpu : This is the queue for CPU jobs. This gives access to 53 compute nodes, each with 16 cores per node.
• gpu : This is the queue for GPU jobs. This gives access to 4 GPU nodes, each with 2 K80 GPUs (which each contain 2 GPU cores, so it looks like 4 GPUs per node).
• all_cluster : This is the queue to allow access to all of the CPU and GPU nodes together. This is a special queue that should only be used when you want to use the whole of BlueGem in a single job. You will need to get permission by emailing Christopher Woods before you can submit a job to this queue.

By default, jobs will go into the "cpu" queue.

Please ensure that you only submit jobs to the "gpu" queue that actually make good use of the GPUs. Submitting CPU-only jobs to the "gpu" queue will block access to the GPUs, preventing them from being used by others.