Gpu utiliaztion factor not working on 301 app

I believe that the rather odd behavior response to a change in the number of GPU tasks differs in timing from the response to a change in the number of simultaneous CPU jobs.

For CPU jobs, one changes computing preferences on the web page, then initiates an update from boingmgr. The moment the update gets through, the revised number of CPU tasks running takes effect on your host.

For GPU jobs, on the other hand, you set the Einstein preference, but running update is not sufficient. Your system must request new GPU work, the server must grant GPU work, and an new entire work unit must be downloaded to your host before the change takes effect.

This does NOT mean that only the newly downloaded work will run with the requested number simultaneously on the GPU--just that the activation of the change on your host somehow does not take effect until that first downloaded WU is fully received.

This might not be an issue in your case, but will be for others, which is my reason for posting it here.