diff --git a/docs/running-jobs.md b/docs/running-jobs.md index c8e1bf24d9762dfce4f6b9196b9f0d0e8d4ee982..409145949149da0dc755b2e9721798b1d7683f34 100644 --- a/docs/running-jobs.md +++ b/docs/running-jobs.md @@ -294,6 +294,13 @@ module load GCC ParaStationMPI srun ./hellompi ``` +:::warning + +Always use the same software stack (e.g. compiler, MPI) that was used to build the software to ensure compatibility and optimal performance. Different versions can cause errors or degrade performance. + +::: + + Remember to specify `gpu:4` `gres` for JUWELS Booster. Then save the script and submit it for execution with: diff --git a/docs/using-gpus.md b/docs/using-gpus.md index efda45dbf2810bf25dd76a0a417c3bfa2ba6cb40..3c634ca8fc915c8fbad9fab1fd9b99d32b55ac78 100644 --- a/docs/using-gpus.md +++ b/docs/using-gpus.md @@ -126,6 +126,12 @@ srun: Job step aborted: Waiting up to 6 seconds for job step to finish. `sgoto` takes the job id as the first argument and the node number within the job as the second argument where the counting starts with 0. `nvidia-smi` prints some useful information about available GPUs on a node, like temperature, memory usage, currently running processes and power consumption. +:::note + +In the example above, `nvidia-smi` shows 0% utilisation because the GPUs are not being used. If GPUs were being used, the utilisation metrics displayed by `nvidia-smi` would be non-zero, reflecting the active use of GPU resources. + +::: + ## GPU Affinity On systems with more than one GPU per node, a choice presents itself - which GPU should be visible to which application task(s)?