I recently setup my SETI-Server (and other equipment) to run Einstein tasks. It currently has a 980 TI and has been crunching all night. The GPU is being used by tasks and everything seems to be fine. However, when running Open Hardware Monitor, it appears GPU core usage is going to 94%, then dropping to 0% and returning back to 94%. It does it in stages too, which are:
On my Gaming Rig, the Pascal GPU seems to do the same thing. Is this normal?
Copyright © 2024 Einstein@Home. All rights reserved.
Yes, it can be. I usually run
)
Yes, it can be. I usually run more than 1 work unit at a time to maintain better GPU usage.
How do you do that? I can't
)
How do you do that? I can't seem to find it in the config.
Michamus wrote:I recently
)
Welcome to the fun house!! :-).
Probably. The apps are quite new and under active development so I don't think we've worked out yet what 'normal' really is :-).
I notice you are running the standard 1.17 version. You should try the new beta test version (1.18) by allowing test apps in your preferences (if you're game and understand the risks - nothing bad seems to happen, though) as it is showing quite a speed improvement from all accounts. Maybe this version might not show as much of the variation you observe.
Cheers,
Gary.
Michamus wrote:How do you do
)
Account -> Preferences -> Project. It's the GPU utilization factor for FGRP apps. You need the same page if you want to run test apps.
Cheers,
Gary.
Gary Roberts wrote:Michamus
)
Change Beta setting to yes then further down under GPU utilization factor change it from 1 to 0.5 then save
Gary Roberts wrote:Welcome
)
Thanks!
Is this in the web preferences? I'll look around and see if I can find it.
EDIT: I just realized you and Zalster set me on the path. Thanks!
You're most welcome!
)
You're most welcome!
EDIT: Just one (or two) further bits of information :-). If you change the GPU utilization factor, clicking 'update' at the client end won't change to multiple tasks. You actually have to download new work for the change to happen. At that point, all tasks in your work cache will be processed using the new factor.
You can also control multiple concurrent tasks irrespective of the website settings through an app_config.xml file. This gives access to finer grained, immediate and individual control of hosts without having to separate hosts into different locations (venues) in order to get different pref settings. It looks a bit difficult when first reading the documentation but it's really trivially easy once you understand the syntax. app_config.xml settings automatically override website settings so once you have finished tweaking you can remove the file and revert to website settings if you so desire.
Cheers,
Gary.