GPU Status Failed

It seems that a lot of GPU problems revolve around specific versions of drivers. Though NVidia has their own support structure, you can often learn from information reported by others who fold.

Moderators: Site Moderators, FAHC Science Team

Burialsheer
Posts: 5
Joined: Sat Aug 17, 2019 10:11 am

Re: GPU Status Failed

Post by Burialsheer »

I tried making FAH use only the Nvidia GPU but still nothing's changed. :/

What the weird thing is that as I explained above I got it to work for a while but after pausing and unpausing it failed again. And it deleted all my GPU progress. I can't imagine why that is.

UPDATE: I now went as far as to disable the Intel GPU completely. Then the GPU switched between "ready" and "running" again for a few times and then surprisingly got to work. However, I tried pausing and unpausing as after doing that earlier it had messed it all up. So I wanted to try if it was fixed for real this time. Aaand it deleted all my progress and does its back and forth switching again that will eventually lead to failing... I'm running out of ideas here honestly.
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: GPU Status Failed

Post by bruce »

When FAH sees a system with ONLY one GPU, here's what happens.

If the GPU happens to be an Intel iGPU, FAH cannot support that GPU which means any WU that has been partially processed will be deleted (and hopefully reported as missing so it can be assigned to somebody else).

If the GPU happens to be a supported GPU (such as NVidia)) a WU will be downloaded and processed.

As long as the system firmware is effectively removing one GPU and replacing it with a different GPU whenever it decides to, you're going have troubles. Somehow you need to disable Optimus's ability to switch GPUs. Apparently disabling the Intel GPU doesn't do that. Are there any other settings (even if they're in the BIOS)?
Burialsheer
Posts: 5
Joined: Sat Aug 17, 2019 10:11 am

Re: GPU Status Failed

Post by Burialsheer »

I haven't looked at BIOS settings yet as I'm not very familiar with that.

In Nvidia control panel I tried setting FAH to only use the Nvidia graphics card + also disabling the Intel GPU. How can my computer still use the Intel one for folding?

Also one other thing I've noticed. If I somehow manage to get my GPU folding then the Run number of the PRCG is always 0. No matter what protein I'm working on or how many times I've tried doing it. Is that normal?

I'm getting rather desperate... A friend of mine has the exact same laptop and for him the GPU folding is working just fine. I can't work out what could be wrong with mine.
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: GPU Status Failed

Post by bruce »

Some project have only one Run. The assignments of Run/Clone numbers depend on some of the factors of the internal construction details of the project and really don't mean anything to us as donors. I consider (Run,Clone) as simply a two dimensional representation of a single concept. As long as you're getting changes to (R,C), you're getting new assignments.

If you want me to look into it further, post some PRCG numbers that you've completed and I'll see if I can dig out any further details.
Burialsheer
Posts: 5
Joined: Sat Aug 17, 2019 10:11 am

Re: GPU Status Failed

Post by Burialsheer »

Okay, now I've seen a few with other Run numbers too so I guess that is random. When I only saw zeros then I suspected that maybe that was an indication that something's up with the GPU etc.

The situation right now is this. I got it pretty much to work if I trod carefully. But then a new Nvidia driver appreared. I installed it and then everything went back to not working again (GPU switches between running and download/ready for a couple of times and then fails). I reverted back to the previous driver but that didn't help. I also again tried deleting the GPU slot and adding it again but that didn't help either.

Has anyone got any other ideas? Should I perhaps send my log again? I'm not an expert so I dunno if there would be something new and helpful to see.

UPDATE: I disabled the Intel GPU again and now GPU folding seems to be able to work on something. The moment I enable Intel again, pause GPU folding and want to start it again, GPU starts to fail. I mean, I could just have Intel disabled but it is still needed for a few things (e.g the screen brightness changing option vanishes if Intel GPU is disabled). I tried enabling-disabling Intel (as I would need to do if I wanted to change screen brightness for example), then pausing and unpausing GPU folding and it luckily kept going. I'm just afraid that maybe once it doesn't work and I'm back to square one with my folding.

UPDATE 2: With Intel disabled, I was pausing and unpausing GPU folding a couple of times. Now doing that deleted all my progress again.
bruce
Posts: 20910
Joined: Thu Nov 29, 2007 10:13 pm
Location: So. Cal.

Re: GPU Status Failed

Post by bruce »

The OpenCL drivers for Intel cause nothing but trouble. Sometimes they are ignored, sometimes not. From FAH's perspective, they are not needed.

When WindowsUpdate installs new drivers, they will REMOVE the OpenCL drivers for NVidia, causing FAH to rapidly discard assigned WUs. Installing drivers directly from NVidia will always work. Unfortunately that means you need to install the updated drivers BEFORE WindowsUpdate decides to do the update for you. (If you're at or ahead of Microsoft's decision to update, you should be OK.)

(It never hurts to submit the first few pages of your log. Sometimes we see things worthy of comment, even if you haven't observed a problem.)
Post Reply