Folding Forum

I have been running two EVGA GTX 780's (stock cards) on separate WinXP machines using Core 17 - every couple of days, both PC's drop out to the Windows log in screen, and on re-entering Windows they report a 'video device driver' error. I've tried all the 320.xx drivers - all have the same issue. I also put one of the cards in a Win7 machine - rather than dropping out, it created bad WU's over the same couple of days time frame.

It is worth noting that all the machine have been Folding, without issue, using GTX 560 cards - indeed, I replaced the GTX 780 with a GTX 560 in one of the XP machines, using the same 320.xx drivers, and there have been no issues in running Core 17.

In the last couple of days - in fact it might be since the core update(?) - the most recent error message changed to 'stop (blue screen) error caused by a device or driver' - this happened on both the WinXP and Win7 machines with GTX 780's.

Am I alone in having these types of problems??

Not saying it's related but I've had very bad problems similar to how you describe with blue screens complaining of the video driver that started almost immediately after a windows update. I've tried updating drivers and going back to a restore point but I think it's a mortal wound needing a re-install of the OS to resolve.

So one check to make is, are they dropping out or rebooting?

Personally, I haven't had much luck with the 320-series drivers. 326.19 beta came out yesterday, give those a try.

Off the top of my head, here are some pointers:
If they are factory overclocked, have you downclocked it to the vendor default?
Do you have enough power for the GPUs?
A clean driver installation also results in the same errors?
Are the systems F@H dedicated or not?
Have you tried a fresh V7 installation or manually updating the FahCore (you delete the FahCore once the GPUs have folded the WUs and then delete the FahCore)?
Are there any wild variations with the GPU frequencies while folding (using MSI Afterburner or EVGA Precision X)?
Are the temperatures within acceptable range?
Is the temperature limit set to 70C or something like that instead of the default of 80C?
Is there something specific running just before the crash or is it completely random?

PantherX wrote:Off the top of my head...
A clean driver installation also results in the same errors?

Check the history log for windowsupdate. Do you see any driver updates?

You probably already know this: They generally put driver updates in the optional category rather than the important category but I suppose if they found a security fault, it could be in either one. (I never use windowsupdate for driver updates but prefer to get them directly from the hardware manufacturer.) Windows does an excellent (??) job of preventing you from ever removing a driver that they installed so if you want to do a clean install, you have to use an add-on program to clean out all traces of the driver you're trying to uninstall.

Thanks for the reply - answers in red below;

PantherX wrote:Off the top of my head, here are some pointers:
If they are factory overclocked, have you downclocked it to the vendor default? Not factory overclock, have also down clocked
Do you have enough power for the GPUs? Yes - 850W (it only has a Q6600 in it)
A clean driver installation also results in the same errors? Yes - did the uninstall/install in Safe Mode, also used Driver Fusion
Are the systems F@H dedicated or not? Yes
Have you tried a fresh V7 installation or manually updating the FahCore (you delete the FahCore once the GPUs have folded the WUs and then delete the FahCore)? Not done a fresh install of v7 since updating the nVidia driver. Core automatically updated recently.
Are there any wild variations with the GPU frequencies while folding (using MSI Afterburner or EVGA Precision X)? Using EVGA Precision - not studied the frequencies
Are the temperatures within acceptable range? Yes
Is the temperature limit set to 70C or something like that instead of the default of 80C? Using the default fan curve in EVGA Precision
Is there something specific running just before the crash or is it completely random? These are normally left Folding without intervention, so I've not been present when it drops out

One thing that was mentioned yesterday was memory and the limits of WinXP32 - given that the GPU has 3GB, and the System 2GB RAM, how would this affect the PC given the limitations on the 32-bit system, i.e. <4GB addressable memory?

bruce wrote:

PantherX wrote:Off the top of my head...
A clean driver installation also results in the same errors? Yes

Check the history log for windowsupdate. Do you see any driver updates? No

You probably already know this: They generally put driver updates in the optional category rather than the important category but I suppose if they found a security fault, it could be in either one. (I never use windowsupdate for driver updates but prefer to get them directly from the hardware manufacturer.) Windows does an excellent (??) job of preventing you from ever removing a driver that they installed so if you want to do a clean install, you have to use an add-on program to clean out all traces of the driver you're trying to uninstall.

I did fresh installs of these drivers following advice from EVGA. In general, I always download from the nVidia driver page.

I have had the 2-day crash in the past and it always was driver related. Reverting or updating the driver always fixed it so far. I have seen discussions about this phenomena in regards to the 32X.XX drivers over in the guru3d forums

In my view it seems to be a driver-related problem - so my guess is to uninstall 32X.XX and install the latest 31X.XX-version.

folding_hoomer wrote:In my view it seems to be a driver-related problem - so my guess is to uninstall 32X.XX and install the latest 31X.XX-version.

I did try this. Unfortunately, nVidia drivers based on 31X.XX won't install with the GTX 780 present - I guess the reason is that only the 320.XX drivers have the added support for the GTX 780.

Same problem with the 326.xx drivers?

7im wrote:Same problem with the 326.xx drivers?

Only installed the 326.XX driver on the XP machine last night - I'll find out if it worked after the two day period ...

Following up on Grendel's comments re: guru3d, it seems like a number of people have been having similar problems with drivers for 700 series cards. Perhaps the 326.XX driver might be the answer (though I won't hold my breath!)

DocJonz wrote:One thing that was mentioned yesterday was memory and the limits of WinXP32 - given that the GPU has 3GB, and the System 2GB RAM, how would this affect the PC given the limitations on the 32-bit system, i.e. <4GB addressable memory?

The 3GB on the GPU is used primarily for raster data when you have high resolution monitor(s) with high color settings. The words I've heard is that FAH uses very little GPU memory though I don't know any numbers that can be associated with that. Since I'm not a gamer, when I have a choice between two GPU models with the same processor chip, I generally choose the one with less video memory. This was certainly adequate for FahCore_11 / 15 / 16. I'm not sure if it has changed much with 17.

Main RAM is a separate issue. The GPU cores need enough memory for a small program and the data associated with the protein. Right now I have Core17 running and it's taking 233 MB but at some point it wanted 436 MB. I also have CoreA3 running using 116 MB. Clearly there's no problem fitting those in almost any system with plenty of space for the OS and for other programs that might be running like this browser than I'm typing in. The SMP cores use more memory as the number of CPUs allocated increase. If you're running a 32-bit OS, it is inadequate for a system with, say 24 CPU cores but with a quad or a hyperthreaded quad like the i7, it should be more than adequate and your 2GB can handle it unless you load it up with a lot of active programs.

Since those systems are dedicated to F@H, how about you do a clean driver installation but only install the driver, not any other features and see what happens. If it still doesn't work, you might want to consider installing Linux (Ubuntu, or Mint are the ones that I have tested and are very easy to configure for F@H) and see if you can get any issues or not.

The 326.49 driver blue-screened with a graphics device driver error this morning .... good job I didn't hold my breath!
I checked the RAM memory while it was Folding - it has 1.3GB spare of the 2GB.

I had tried to get one of my Ubuntu boxes working with 320 drivers to test it, but was having some issues getting the drivers in (I dabble with Linux, but am no expert ....) - so, I've just re-conjured the Linux machine to run Win7x64. I have put a GTX 560 into this machines with the 320.49 (latest WHQL) drivers and left it Folding (GPU only). I'll monitor this over the coming days, and if everything is fine, I'll swap out the GTX 580 for a GTX 780 ....