HPC in CoE (College of Engineering)

Contributors: Brodie Pearson

Access this resource via the HPC Dashboard or through standard SSH [NOT WORKING]:

ssh [username]@submit-a.hpc.engr.oregonstate.edu

This resource can be used similar to the more extensive CQLS documentation page. But there is some specific CoE-specific information below.

Using the preempt partition

A command like this will start an interactive session on a V100 which is suitable for Oceananigans simulations. The preempt partition allows users to request communal resources that are in low demand:

srun -p preempt --gres=gpu:1 --constraint=v100 --pty bash

So far we have had success running Oceananigans on the NVIDIA V100 GPU (see above constraint) and on the RTX6000 (replace with --constraint=rtx6000)

Advanced GPUs for future research

If GPUs are useful for your research going forward, you can also access the new dgxs or dgxh nodes via CEOAS or Engineering group (may need to replace dgxs with dgxh or dgx2 depending on your group permissions), for example:

srun -p dgxs --gres=gpu:1 --mem=8G -t 00:10:00 -N 1 --pty /bin/bash