About GPU Resources in O2
The first 2 3 GPU nodes are now available on O2, including: 4 Tesla M40 and 8 16 Tesla K80 GPUs. To list information about all the nodes with GPU resources you can use the command:
login01:~$ sinfo --Format=nodehost,cpusstate,memory,gres|grep 'HOSTNAMES\|gpu' HOSTNAMES CPUS(A/I/O/T) MEMORY GRES compute-g-16-177 0/0/24/24 257548 gpu:teslaK80:8 compute-g-16-176 108/1012/0/20 257548 gpu:teslaM40:4 compute-g-16-177194 16/2314/0/2420 257548 gpu:teslaK80:8
GPU Partition Limits
The following limits are applied to this partition in order to facilitate a fair use of the limited resources:
The amount of GPU resources that can be used by each user at any time in the O2 cluster is measured in term terms of GPU hours / user, currently there is an active limit of 72 GPU hours for each user.
This means that For example at any time each user can allocate* at most 1 GPU card for 72 hours , or 12 GPU cards for 6 hours or any intermediate combination, for example 6 GPU cards for 12 hoursother combination that does not exceed the total GPU hours limit.
The current limit will be increased * as resources allow
The total amount of memory, from all running GPU jobs, that each user can get allocated is set to 250GB
The total amount of CPU cores, from all running GPU jobs, that each user can get allocated is set to 20
Those limits will be adjusted as we migrate additional GPU nodes from the older cluster to O2.
* as resources allow
How to compile cuda programs
To submit a GPU job in O2 you will need to use the partition gpu and must add the flag --gres=gpu:1 to request a GPU resource. The example below shows how to start an interactive bash job requesting 1 CPU core and 1 GPU card: