GPU
General GPU servers
Do not use GPU servers unless you are running jobs on the GPUs. If your jobs will run on the CPU you should be using the Compute servers.
We have a few GPU servers that are available for general access. These are available for deep/machine learning with python libraries, or running Matlab code directly on the GPUs.
Most researchers use python virtual environments to run machine/deep learning programs on these servers. I recommend that you read our guide on how to setup python for deep/machine learning. Please contact us if you need any assistance with this.
The servers that you can use are:
| HOSTNAME | CPU | RAM | GPU's |
|---|---|---|---|
| medusa.ee.ucl.ac.uk | 2 x Intel(R) Xeon(R) Gold 5120 CPU @ 2.20GHz | 128Gb | 2 x NVidia 16Gb Tesla V100 |
| cork.ee.ucl.ac.uk | 2 x Intel(R) Xeon(R) Gold 5120 CPU @ 2.20GHz | 256Gb | 2 x NVidia 16Gb Tesla V100 |
| athens.ee.ucl.ac.uk | 4 x Intel(R) Xeon(R) Gold 6126 CPU @ 2.60GHz | 512Gb | 4 x NVidia 32Gb Tesla V100 |
| london.ee.ucl.ac.uk | 2 x Intel(R) Xeon(R) Gold 6248 CPU @ 2.50GHz | 512Gb | 10 x NVidia 32Gb Tesla V100 |
| geneva.ee.ucl.ac.uk | 2 x AMD EPYC 7543 32-Core CPU | 512Gb | 4 x NVidia 80Gb Tesla A100 |
| turin.ee.ucl.ac.uk | 4 x Intel(R) Xeon(R) Gold 6242 @ 2.80GHz | 384Gb | 4 x NVidia 80Gb Tesla A100 |
Info These resources can get very busy. There are also many other HPC resources available to you for free, run by UCL ARC. These include CPU and GPU processing services. Please get in touch with ARC, or speak to the departmental IT team about how to make use of these services.
WarningThese servers can only be accessed from within the UCL network. If you are working remotely you will either need to connect to the UCL VPN, or connect to our departmental SSH Gateway first (ssh.ee.ucl.ac.uk).
Errors If you are having problems using the GPU servers, take a look at our GPU debugging guide