The Research Computing department offers specialized servers equipped with GPUs that can perform many calculations at the same time. These GPU servers are ideal for jobs that involve large amounts of data or complex computations, such as deep learning, molecular modeling, or image processing. By using a GPU, users can often complete jobs much faster than on regular CPU nodes. This allows users to test ideas more quickly, run larger models, and get results in less time.
What GPUs do I have access to?
Users can list the GPUs they have access to by executing the following command:
[user@rocky-login012 ~]$ sinfo -o "%P %N %G"
PARTITION NODELIST GRES
shortq7* nodegpu022 gpu:p100:4
shortq7* nodegpu024 gpu:titanx:3
shortq7-gpu nodegpu010 gpu:rtx3070ti:1
shortq7-gpu nodegpu[039-041] gpu:a100:3
shortq7-gpu nodegpu[102,104-116] gpu:v100:4
How do I search for a certain GPU?
Use | grep -i GPU_NAME to search for a specific GPU:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i a100
longq7-rna nodegpu003 gpu:a100:3
longq7-soa nodegpu042 gpu:a100:2
longq7-eng nodegpu[039-041] gpu:a100:3
shortq7* nodegpu[003,039-041] gpu:a100:3
shortq7* nodegpu042 gpu:a100:2
shortq7-gpu nodegpu[039-041] gpu:a100:3
interactive nodegpu[039-041] gpu:a100:3
What GPUs are available in a certain partition?
Use -p PARTITION_NAME to list all the GPUs available in a specific partition:
Note: Certain partitions are available only to students, staff, and faculty if they are from the respective school or group.
[user@rocky-login012 ~]$ sinfo -o "%P %N %G" -p longq7-eng
PARTITION NODELIST GRES
longq7-eng nodegpu[039-041] gpu:a100:3
longq7-eng nodeamd[009-017,021-029,031,033-035,038],nodeintel[001-004,006,009-010] (null)
How do I use a GPU on the cluster?
sbatch
See the following article for creating an sbatch script:
https://helpdesk.fau.edu/TDClient/2061/Portal/KB/ArticleDet?ID=141487
You may include either of these lines in your sbatch script to request a GPU node.
Request any GPU:
#SBATCH --gres=gpu:1
The number at the end can be replaced to request more GPUs. e.g., gpu:2, gpu:3
Request a specific GPU:
#SBATCH --gres=gpu:a4000:1
The number at the end can be replaced to request more GPUs. e.g., a4000:2
salloc
See the following article for salloc:
https://helpdesk.fau.edu/TDClient/2061/Portal/KB/ArticleDet?ID=141472
[user@rocky-login012 slurm]$ salloc -N 1 --gres=gpu:a100:1
salloc: Granted job allocation 1426376
Thu Oct 30 14:41:26 EDT 2025 Starting spack.sh
Thu Oct 30 14:41:27 EDT 2025 Complete spack.sh
[user@rocky-login012 slurm]$ squeue -u $USER
JOBID PARTITION NAME USER ST TIME NODES NODELIST(REASON)
3482136 shortq7 interact user R 0:07 1 nodegpu039
[user@rocky-login012 slurm]$ ssh nodegpu039
Last login: Thu Oct 30 18:20:08 2025 from 10.116.19.32
Thu Oct 30 18:41:38 UTC 2025 Starting spack.sh
Thu Oct 30 18:41:39 UTC 2025 Complete spack.sh
[user@nodegpu039 ~]$
List of all GPUs on the Athene cluster (current as of October 30, 2025):
V100:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i v100
longq7-mri nodegpu[102,104-116] gpu:v100:4
longq7-video nodegpu[114-116] gpu:v100:4
shortq7* nodegpu[102,104-116] gpu:v100:4
shortq7-gpu nodegpu[102,104-116] gpu:v100:4
A100:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i a100
longq7-rna nodegpu003 gpu:a100:3
longq7-caai ca-ai gpu:a100:3
longq7-soa nodegpu042 gpu:a100:2
longq7-eng nodegpu[039-041] gpu:a100:3
shortq7* nodegpu[003,039-041] gpu:a100:3
shortq7* nodegpu042 gpu:a100:2
shortq7-gpu nodegpu[039-041] gpu:a100:3
interactive nodegpu[039-041] gpu:a100:3
node_test nodegpu040 gpu:a100:3
P100:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i p100
longq7-rna nodegpu022 gpu:p100:4
shortq7* nodegpu022 gpu:p100:4
Titan X:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i titanx
longq7-rna nodegpu024 gpu:titanx:3
longq7-rna nodegpu[023,025] gpu:titanx:4
shortq7* nodegpu024 gpu:titanx:3
shortq7* nodegpu[023,025] gpu:titanx:4
A4000:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i a4000
longq7-rna nodegpu[026-028] gpu:a4000:4
shortq7* nodegpu[026-028] gpu:a4000:4
shortq7-gpu nodegpu[026-028] gpu:a4000:4
A5000:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i a5000
longq7-rna nodegpu[029-035] gpu:a5000:4
shortq7* nodegpu[029-035] gpu:a5000:4
shortq7-gpu nodegpu[029-035] gpu:a5000:4
RTX 3070 Ti:
[user@rocky-login012 ~]# sinfo -o "%P %N %G" | grep -i rtx3070ti
longq7-rna nodegpu010 gpu:rtx3070ti:1
shortq7* nodegpu010 gpu:rtx3070ti:1
shortq7-gpu nodegpu010 gpu:rtx3070ti:1