Page 3 of 3
Re: Enormous job: how did I get it, how can I fix it?
Posted: Mon Oct 05, 2020 2:52 pm
by bruce
Be careful. You cannot tell if a WU is an "enormous job" until FAHClient has processed enough of that project to figure out how fast your system is. The initial estimate is a wild guess but as soon as some progress has been reported, it converges on a realistic estimate. If you have a history of having worked on that particular project, it will use that information... if not, you'll have to wait until a checkpoint has been completed.
I think it might use the timeout as the initial estimate.
How many hours per day is FAH running? ... and is FAH set to run only-when-idle?
Re: Enormous job: how did I get it, how can I fix it?
Posted: Mon Oct 05, 2020 4:11 pm
by Joe_H
bruce wrote:I think it might use the timeout as the initial estimate.
Yes, currently the timeout values are used for initial estimates on any project not folded before.
Re: Enormous job: how did I get it, how can I fix it?
Posted: Mon Oct 05, 2020 6:00 pm
by bruce
... which generally overestimates the duration of new projects.
Re: Enormous job: how did I get it, how can I fix it?
Posted: Tue Oct 06, 2020 1:29 am
by midhart90
PantherX wrote:Welcome to the F@H Forum midhart90,
If you can describe your hardware and use-case, we might be able to discuss this with the researcher and see what we can do.
I have a Radeon HD 7700 connected to an older, but still serviceable, Windows 10 machine. Whenever I get a WU with a 1-day timeout, it seems that it always takes just a bit over a day to finish and ends up getting reassigned. Yet, when I get one with a 2-day timeout, in my experience, it does not take significantly longer to fold, and therefore finishes on time. Is the timeout duration supposed to be proportional to the size of the WU?
Re: Enormous job: how did I get it, how can I fix it?
Posted: Tue Oct 06, 2020 3:47 am
by Neil-B
Not necessarily .. as I understand it, timeout (and expiration) is set by the researchers on various factors .. with covid/moonshot projects there has been an urgency for results to be returned quickly and so timeouts have been shorter than traditional projects ... you may therefore see some WUs with shorter timeouts than others that actually take similar processing times.
Re: Enormous job: how did I get it, how can I fix it?
Posted: Tue Oct 06, 2020 4:15 am
by psaam0001
I do not know if this will help, but it may be worth a try since you are running Windows. If you know how to get to the settings for power saving, set it to maximum performance. Then click the link for advanced settings and look for PCI Express. Double click on that line and click on link state management. Then set it to off before clicking on OK.
Paul
Re: Enormous job: how did I get it, how can I fix it?
Posted: Thu Oct 08, 2020 2:14 am
by midhart90
psaam0001 wrote:I do not know if this will help, but it may be worth a try since you are running Windows. If you know how to get to the settings for power saving, set it to maximum performance. Then click the link for advanced settings and look for PCI Express. Double click on that line and click on link state management. Then set it to off before clicking on OK.
Paul
I checked this and link state management was already disabled. That being said, my GPU appears to be getting dealt appropriately sized WUs now, so it appears this is resolved (at least for now). Thanks for the help!
Re: Enormous job: how did I get it, how can I fix it?
Posted: Thu Oct 08, 2020 5:36 am
by psaam0001
midhart90 wrote:I checked this and link state management was already disabled. That being said, my GPU appears to be getting dealt appropriately sized WUs now, so it appears this is resolved (at least for now). Thanks for the help!
No problem. I just brought that up, as I can't see what your settings were, without doing a Microsoft Teams or Teamview session.
Paul