Nvidia Windows Kernel Mode Driver stopped and recovered

ylweyez

Honorable
Sep 1, 2013
4
0
10,510
I've looked around the internet and found a few helpful hints, but I am still having the following problem with my EVGA GTX 650 Ti Boost

I guess I will start by saying I just built this computer a week ago, and have been noticing more and more that my screen goes black, and then turns back on to a windows bubble that says:

Display drive stopped responding and has recovered
Display driver Nvidia Windows Kernel Mode Driver, version 320.49 stopped responding and successfully recovered.

This also happens randomly. I have played several games,Elite Sniper 2 and Team Fortress 2, for several hours without a problem. It seems to happen more when I switch a tab in chrome, or try to open a folder to view it.

Some things I have tried:

-Reinstalling the Nvidia Drivers direct from their site (using 320.49 now as the most recent stable version)

-Making sure my motherboard drivers were up to date.

-A forum suggestion was to turn up the power 50 mV via precision which I did.
*This seemed to work for half a day, and then I had 3 instances of the black screen in a row, so I turned it back down.

My concern is that I only have a limited window to return the graphics card if that may be the problem.

I am using a USB Adapter for WiFi, I have also noticed that sometimes on startup the driver fails to load and I have to restart the computer or unplug and replug in the device. Doubt it affects the graphics card at all but it may point to a problem somewhere else?

That is why I am turning here. I'm at a loss. Thanks for helping.

Stats:
Gigabyte GA-990FXA-UD3 Motherboard
AMD FX-8120 3.1GHz
EVGA GTX650 TI Boost
750W PSU
Kingston Hyper-X Black Edition 4GB x 2
Western Digital Blue 500 GB 7200 RPM HDD
Windows 7 Home Edition 64-bit

Hooked up DVI-D to an AOC 69V 22" monitor
 
Solution
Well that went much farther than my old card did.

I have been getting some assign-GPU.stanford.edu:80 errors, I just figured it was my isp. still gets a WU after a few tries.

I am not 100% sure if the power settings would make a difference or not.

If F@H does not crash

1. I hope it is just a driver issue that may be fixed in the future.
2. F@H Presents a high enough load that the card runs into its own power limit and does not clock up as much as when playing games. My 650 ti is NOT a boost model so it just goes all in when folding. My 670 gets near top clocks when folding.

If you want to see how high the card is clocking, You would need to use a tool like GPU-Z to log these stats in the background and then a program like Excel or...
My only suggestion to test this would be to find a way to crash the card on command. This way you can try things and see quickly if it fixes it.

One thing you can try is Folding @ Home(disable the cpu slot and just use the GPU slot.)

Now see if that produces an unstable machine error in the logs(it also restarts what it was doing, so you will be able to tell if it has had an issue.). Please note this does NOT crash the driver as it is running a CUDA program on the card and not a game or video.

If you DO get this issue(unstable machine error), clock down the card by about -39mhz. What I am trying to do is find out at what point the card will be stable. Going 13 at a time is best as the Nvidia boost works in increments of 13.

This is what the failing log looks like
7095.png


I know this seems like allot of work, but I find F@H crashes faster than ANY game on an unstable card.

It also does not hurt to keep an eye on the temps while doing this.

This is a log of temps of a crashing GTX670 running F@H. The new one does not seem to crash any more(has gone for days on end), but time will tell.
km7.png


I started looking for a faster way to track Nvidia card issues and this was the one I came up with.
 

ylweyez

Honorable
Sep 1, 2013
4
0
10,510
Is the Folding @ Home the protein simulator by Stanford? If it is when I run that I don't know how to disable the CPU, and the graphics card is idle, and says "Wait X minutes" and then when it counts down it just repeats to a higher x value. Not sure if I'm on the right path or not.
 
Setting Folding @ Home to run on your GPU(Video card ONLY)

1. Right click the F@H Icon and select Advanced Control(You can also set Folding to Off while making these changes.)
emrn.png

2. You should get a new window with a Configure option
3. Select the Slots tab
4. Select CPU(Your cpu may be in a different slot than mine, you just want to remove cpu) and than use the remove button
5. Save
vfn4.png


You can now slide the Folding Power slider to Full and the video card should start working(it take a short time to download a work unit for the card to perform).

If you select the Logs tab after folding has started you can clear the logs(just re remove any logs generated from the changes you made) and check Warnings and Errors so you can more easily see the Unstable Machine error if you get on.

Now if you DO start to get those, use EVGA's software of MSI After Burner(do not use both together) to drop the clock speed at least 39 mhz(if you want to go past this add 13 per step).

If you find a place that can fold without crashing. You can try some games. You will HAVE to stop folding @ home or set the slider to anything lower than FULL(other settings will only take place when the system is idle) before playing games or they will run like crap.

A side effect of having this installed is also that the computer will NOT be able to go to sleep. If you tell it to, it may just shut the screen off, but will leave the computer on.

Now if you like the idea of helping Stanford, you can run F@H on idle time for team 40051 :)
 

ylweyez

Honorable
Sep 1, 2013
4
0
10,510
Well I ran this for a while today with no problem. I watched on Precision my temperature of the card rise to about 60-61 degrees C and my fan follow up the rise by going from 10% to 40% power. The screen saver even came on and you wouldn't have known anything was running at all.

I then shut the program off and put the computer to sleep while I went away for probably no more than 3 hours, and then when I woke the computer up I checked the logs and I had another kernel error message about the second I woke the computer up. It just doesn't make sense. It's been fine since then though...

I did read one more solution that went into changing the power management settings for the card from power saver modes to High performance balls to the wall performance. I'm not saying its working, but it seems to be at the moment.

The only other error I'm reading from the event log is a DNS failure.

But I will continue to run the program for team 40051 for all your help.
 
Well that went much farther than my old card did.

I have been getting some assign-GPU.stanford.edu:80 errors, I just figured it was my isp. still gets a WU after a few tries.

I am not 100% sure if the power settings would make a difference or not.

If F@H does not crash

1. I hope it is just a driver issue that may be fixed in the future.
2. F@H Presents a high enough load that the card runs into its own power limit and does not clock up as much as when playing games. My 650 ti is NOT a boost model so it just goes all in when folding. My 670 gets near top clocks when folding.

If you want to see how high the card is clocking, You would need to use a tool like GPU-Z to log these stats in the background and then a program like Excel or Open Office Calc to turn this data into a chart.

I hate to say it, but it can take time to test this kind of thing. In general, If a game averages high clock speeds and causes problems while games that average lower clock speeds does not, chances are dropping the clock a bit may help. If you can confirm that IS the issue, the card should be able to be RMA'd for sure.

The game RAGE on top vs Bioshock Infinite on the bottom. Rage is so easy to play the card never needs to even clock up.
afeg.png
 
Solution