Folding@Home hangs my system since update

[Ozaron]

Hi All, first thread on the forums, go easy on me!

So as of roughly a week ago, attempting to run F@H workloads on my GPU causes my system to lock up. Within a few seconds of starting, I lose any keyboard or mouse input, and disconnecting either peripheral and reconnecting it leads to a situation where they don't even power on, as if the USB ports aren't functioning anymore. The screen stays active, displaying the last frame it was sent indefinitely, as far as I understand. The only way to end its misery is power button.

I believe this began after a Windows 10 update (which I had put off for a few weeks but caved in and allowed when I had time to let my poor internet do its thing). I had previously been running the 17.3.3 drivers for Radeon Crimson or whatever they call it nowadays. Windows of course "updated" me to 17.1.1, after which I had a different issue whereby the driver would crash and recover every 30 minutes or so. I fixed that by updating to 17.5.2, and can confirm that after a straight 9ish hours of use including in stuff like 1440p Overwatch, I have no issues.

But on returning to F@H? Still locks up.

I haven't completely wiped my drivers and started again due to time consumption (please be kind, the drivers are over 500mb total and my 'net tops at 130kb/s right now.. ) but I'd love to hear any input. Considering games and desktop now work OK I'm sort of loathe to wipe drivers just for F@H but will do it if needs must.

System description is accurate, R9 Fury running at 1065MHz core 512MHz HBM. Never had any issues with it (over HDMI at least) and it can clock higher.

Thanks in advance.

[jimbouk]

Someone will turn up with the recommendation for a good driver cleaning program, but that would definitely be my suggestion. Can you download the drivers elsewhere and pen-drive them over?

Took me a while to get things working again after the last time I updated my graphics drives. Took a few uninstalls, restarts, clean-ups, re-installs to get me stable again, should have used something to help but couldn't remembers the program and was hoping to play online games with people at the time!

[DDY]

The lack of user control over drivers is one of the reasons I gave up Win 10.

The only method that worked for me, at least temporarily and for some drivers, is to run the update hider tool;

https://support.microsoft.com/en-us/...-in-windows-10

Not all drivers appear on the list and eventually the drivers will be updated, be it because the time limit to defer updates runs out, or you're forced to update the version of Win 10, or whatever MS decides to do this month. It's a pain in the ass having to do this over and over again, but it's the only method that I know of to prevent drivers from automatically updating.

EDIT: Another thing springs to mind, perhaps a bad batch of work units have been released? I had issues in the past with certain GPU WUs locking up my machine. Once the F@H project stopped handing out those bad WUs, the locking up problem would stop.

Have you noted which WUs your machine works on when it hangs?

[Ozaron]

Can you download the drivers elsewhere and pen-drive them over?

Will do that today, hopefully work won't mind...

Took me a while to get things working again after the last time I updated my graphics drivers. Took a few uninstalls, restarts, clean-ups, re-installs to get me stable again, should have used something to help but couldn't remembers the program and was hoping to play online games with people at the time!

Well considering the worst that has happened since re-updating to 17.5.2 is Relive being turned off and F@H being allergic to proteins, I think this has gone fairly well. Still remember my last AMD GPU setup would cause a BSOD every time I ran the driver installer..

The lack of user control over drivers is one of the reasons I gave up Win 10.

The only method that worked for me, at least temporarily and for some drivers, is to run the update hider tool;

https://support.microsoft.com/en-us/...-in-windows-10

Not all drivers appear on the list and eventually the drivers will be updated, be it because the time limit to defer updates runs out, or you're forced to update the version of Win 10, or whatever MS decides to do this month. It's a pain in the ass having to do this over and over again, but it's the only method that I know of to prevent drivers from automatically updating.

Shall check this out shortly. I knew that I wouldn't be able to avoid the driver installation when I saw it in the list. I'm sure people frequently have issues with drivers when they don't update but there's no reason not to allow (especially on Enterprise!) an off button for driver installs for the digitally literate. It's not surprising you and others want to leave when Microsoft has a case of "lowest denominators" coming on.

EDIT: Another thing springs to mind, perhaps a bad batch of work units have been released? I had issues in the past with certain GPU WUs locking up my machine. Once the F@H project stopped handing out those bad WUs, the locking up problem would stop.

Have you noted which WUs your machine works on when it hangs?

This is where I think the answer lies, actually. I haven't got the last unit written down anywhere and I can't check it from work but my theory is that I may have had F@H running when the reverse-update to 17.1.1 happened, and when the drivers crashed it could have corrupted or damaged the unit? That would explain to me why the system lockup is consistently happening when I try to begin folding on GPU. The same unit is still loaded in to my client, and I believe it's at around 17% progress.

There's an option to flush a work unit (not a euphemism) somewhere isn't there? Otherwise I have to wait for it to time out.

Anyway, cheers folks. I'll fiddle with it later today.

[DDY]

To delete a WU; open up your start menu and navigate to the F@H folder and open up the "Data directory" shortcut.

There should be at least one folder in there, possibly more than one, their names (well, number) corresponds to the WUs in your F@H work queue. Identify which one is the problematic WU (i.e. the one that the GPU is stuck on) and delete that folder. You can see the work queue in the F@H client advanced control.

[Ozaron]

Alright, so...

Tried opening up advanced control for F@H as usual and couldn't, given that it was impossible to pause when I last had the system lock up. Immediately it begins folding and happens again, and I couldn't get into the control panel.
Followed your instructions to delete the work unit from the "work" folder in data directory. Re-opened F@H advanced control, let it download a new work unit, 0.00% complete before system locks up again. So, it's not the unit itself. (Or wasn't.)

Immediately fell back on the previous plans, and uninstalled both F@H and my Crimson drivers. Uninstalled Crimson via its own uninstaller, and rebooted. Installed the newest driver they have for R9, rebooted. Now running 17.6.1. At this point something unhappy occurs - the system locks up while I don't even have F@H installed. Figuring it's no longer even the software, nor the GPU drivers, I return to BIOS, reset pretty much everything to default, boot back to Windows, reset my GPU voltage modifier to default, install F@H again (and give it a more accessible data folder than roaming) and let it go. Now it's folding on both CPU and GPU just fine, though I haven't ramped up to Full yet.

So it's probably just my system config, somehow. Might run a couple of tests on memory soon...

Thanks for your help, I'll update if things get worse.