SOLVED Something is destroying my folding. Bad GPU or RAM?

Last night I was using my computer and I noticed my 8800GT had stalled. It didn't EUE, just slowed to a crawl, and my linux vm had dropped about 15k points, so I checked task manager to find "System" taking up 13% of CPU (a full core). I opened Process Explorer and traced it back to dxgmms1.sys. Google revealed it's a common problem with overclocking for it to be found as a "cause" for a BSoD, but I wasn't having a BSoD. After hours of research, I decided to reboot (I hate to reboot THAT much). It was gone for a few minutes. Fired up the GPU folding and let it set. It was fine. Ran the linux vm and it was fine for a little bit more, then happened again.

Next I backed off the overclock on the cards and tried again. This time it was EVGAPrecision.exe taking up a lot of CPU. I disabled EVGA Precision and tried again. That time folding would slow to a crawl and the display locked up. The odd part is that my 9800GT is driving the monitor while the 8800GT was the one seeming to have problems. I finally decided to run MemtestG80 on the 8800GT and it bluescreened.

Then I backed off my CPU and RAM overclocks (well, underclocked my RAM) and tried memtestG80 again. Display locked up. But I could remote desktop into it fine. I updated drivers because I was getting desperate, then tried again. Same results. Then I decided to not use GPU folding at all and ran just the VM over remote desktop. It ran for about 10 minutes then my remote desktop started dying. Turned on the monitor to find a locked up display. Backed off CPU OC even more and ran it again for about 15 minutes, then went to bed. Woke up and found it locked up again.

I'm now at stock and it's been running long enough for me to type this with a linux VM. Obviously, stock is unacceptable . I also want to get at least 1 GPU going again.

So basically, I have no idea what this is symptom of. The locked up display makes me think it's a GPU problem, but it seems to be affecting both GPUs, even when the GPUs are completely idle. So my only other thought is RAM, however I realllllly don't want my main computer down for hours and hours while running a memory test unless it's almost definitely RAM. I'm really frustrated here, cause I have no errors of any kind other than the high CPU usage and a bluescreen pointing to an Nvidia file.

I had a problem with the rpocess "system" eating a lot of cpu cycles with a 295 GTX running the 266.xx drivers. I reverted to 258.xx and it went away. If you're running this on the 2600K, you don't need the gpu to fold anyway. It's just a waste of electricity.

It gives me another 4k or so. :-\ However, whatever it is seems to be affecting my CPU folding too. Even with the GPUs idle it locks up my computer and I have to hard power off. Even if I'm RDPed in, "shutdown /r /t 0" shuts down everything, but won't reboot.

And for some reason the Linux VM doesn't seem to take off from where it left. It starts a whole new WU. So unless I can make the deadline with no OC with no lock ups, I'm down for CPU folding too.

Try Uninstalling Precision it was acting up on me and I removed it. Then I would check your last restore points and see if any windows update or something coincides with the problem and revert back. See if that helps.

dfonda said:

Try Uninstalling Precision it was acting up on me and I removed it. Then I would check your last restore points and see if any windows update or something coincides with the problem and revert back. See if that helps.

Click to expand...

The odd thing is that this happened when NOTHING changed. It's just been sitting folding for days and then this happened. No new software, hardware, drivers, config changes, updates, etc. That makes me think / worry that I've just been pushing the GPUs too hard too long. I also considered my PSU going, since I added 2 more internal hard drives and a USB3 hard drive last week. But it still happens with one GPU not folding, so it's not THAT much power. Plus my PSU should be good.

I found this http://support.microsoft.com/kb/983615 (a hotfix for the exact problem I'm having) though, which I'm really hoping will work. I have no idea why that would have suddenly started though

Crap, I just realized that it probably won't fix the lockups with GPUs idle. Or maybe it will.

PeddlerOfFlesh said:

And for some reason the Linux VM doesn't seem to take off from where it left. It starts a whole new WU. So unless I can make the deadline with no OC with no lock ups, I'm down for CPU folding too.

Click to expand...

This is totally wrong, if you are using linuxrouters image, it write a new image of my work every time I shut down vmware and restores it to the exact point where I quit. It is a fairly long write to the hard drive, you might have a hard drive that is dieing.

spitter3 said:

This is totally wrong, if you are using linuxrouters image, it write a new image of my work every time I shut down vmware and restores it to the exact point where I quit. It is a fairly long write to the hard drive, you might have a hard drive that is dieing.

Click to expand...

It does if I suspend it.

Anyway, after more and more testing (hotfix didn't fix it) I decided to run OCCT instead of folding, so I won't keep screwing up WUs. I noticed my 8800GT was at 80C at idle. Turned the fan up to 80% and it stayed there. Went and took the card out and could barely move the fan with my fingers. I'm willing to bet $1 that this overheating GPU was the culprit. Not sure why it'd take down my other card / display though.

Restored my CPU and GPU overclock and it's been working fine for about 45 minutes.

Up untill the time your system started slowing down, did you hear any noise like a grinding noise?, If you do, and you get to the fan before it locks up, you can put a drop of very light oil (like for sweing mashine oil) on the bearings of the fan (you will need to pull back a lable to do this) you can save the fan and extend it's life.

PeddlerOfFlesh said:

It does if I suspend it.

Anyway, after more and more testing (hotfix didn't fix it) I decided to run OCCT instead of folding, so I won't keep screwing up WUs. I noticed my 8800GT was at 80C at idle. Turned the fan up to 80% and it stayed there. Went and took the card out and could barely move the fan with my fingers. I'm willing to bet $1 that this overheating GPU was the culprit. Not sure why it'd take down my other card / display though.

Restored my CPU and GPU overclock and it's been working fine for about 45 minutes.

Click to expand...

WhitehawkEQ said:

Up untill the time your system started slowing down, did you hear any noise like a grinding noise?, If you do, and you get to the fan before it locks up, you can put a drop of very light oil (like for sweing mashine oil) on the bearings of the fan (you will need to pull back a lable to do this) you can save the fan and extend it's life.

Click to expand...

Yeah, I might have heard it, but I'm almost never physically near the computer. It's kinda an HTPC. It's just in the office with wires going through the wall to a TV and receiver and then I use it for every day stuff with remote desktop. I e-mailed PNY and they said they'd RMA it if it's under warranty (it might be, if the warranty is 3 years) otherwise I might rig another fan on, or get a waterblock. I don't know. Switching to a linux VM and getting my RAM to 1600MHz @ 8-8-8-24 and only one GPU has given me an extra 10k ppd. Will be nice in the summer. The house was noticeably colder this morning, like it always is when GPU folding goes down.

I've been stable for hours now, so I'm guessing that was the problem :-\. Man, all the hours I could have saved if I had just noticed the GPU temp earlier. And that should have been the FIRST thing I checked!

If you can remove the fan, see if putting any oil in the fan will free it up? If not oh well good try

PeddlerOfFlesh said:

Yeah, I might have heard it, but I'm almost never physically near the computer. It's kinda an HTPC. It's just in the office with wires going through the wall to a TV and receiver and then I use it for every day stuff with remote desktop. I e-mailed PNY and they said they'd RMA it if it's under warranty (it might be, if the warranty is 3 years) otherwise I might rig another fan on, or get a waterblock. I don't know. Switching to a linux VM and getting my RAM to 1600MHz @ 8-8-8-24 and only one GPU has given me an extra 10k ppd. Will be nice in the summer. The house was noticeably colder this morning, like it always is when GPU folding goes down.

I've been stable for hours now, so I'm guessing that was the problem :-\. Man, all the hours I could have saved if I had just noticed the GPU temp earlier. And that should have been the FIRST thing I checked!

Click to expand...