Guide: Queuing Directly to a GPU
How to queue jobs directly to a specific GPU, overriding the built-in job scheduler.
Last updated
How to queue jobs directly to a specific GPU, overriding the built-in job scheduler.
Last updated
As of CryoSPARC v2.12+, it is possible to easily queue jobs directly to a specific GPU, overriding the built-in job scheduler. This feature is useful for advanced users that have direct control over their compute resources, and would like to push the limits of their GPU utilization (assuming the GPUs have the required memory available).
To use the feature, build and queue a job as you normally would. You will be met with the new queue page.
You can queue the job to a lane as you normally would, or, you can queue to a GPU by clicking on the "Run on specific GPU" tab.
Select the GPU(s) you would like to run your job on. You cannot assign a job to GPUs across different nodes.
Hit Create.
The job will now be executed immediately on the machine you've selected, regardless of whether the machine has enough CPU or RAM resources available.
