Video driver crash

Hi,

I am also troubled by the video driver crash problem more people seem to suffer from. I'm running on a MSI GTX 670 and somewhere within playing for half an hour the screen goes black. Sometimes the driver resets itself but most of the times I have to reboot my computer. Most of the times there is nothing to be found in the windows event logger and also nothing in the DCS log files.

I monitored the temperature of my GPU but that does not seem to be the problem, it does not het hotter than about 70 degrees and sometimes it happens imedialtely after the game starts.

I already tried everything I could find in this forum (complete reinstall, set some windows timeout registry key, changed some settings).

I'm now trying to revert back to older drivers. I tried the 377.88 driver but that didn't solve the problem.

Is there anything else I could try (or there a certain driver version that is gueranteed to work?)?

Thanks, Marcel

Hi...have had the odd crash lately...dont know if its video related.Only had them in last 2 builds of dcs world..ie 1.28 was fine

Try nvidia v331.82

Try going to bios and changing the pci express setting to Gen 2 from auto. Be under miscellanous settings if its a gigabyte board. Had this exact problem with a inno3d 670. Guessing youre using Windows 8.

Hi,

 

I am also troubled by the video driver crash problem more people seem to suffer from. I'm running on a MSI GTX 670 and somewhere within playing for half an hour the screen goes black. Sometimes the driver resets itself but most of the times I have to reboot my computer. Most of the times there is nothing to be found in the windows event logger and also nothing in the DCS log files.

 

I monitored the temperature of my GPU but that does not seem to be the problem, it does not het hotter than about 70 degrees and sometimes it happens imedialtely after the game starts.

I already tried everything I could find in this forum (complete reinstall, set some windows timeout registry key, changed some settings).

I'm now trying to revert back to older drivers. I tried the 377.88 driver but that didn't solve the problem.

Is there anything else I could try (or there a certain driver version that is gueranteed to work?)?

 

Thanks, Marcel

I know you have a msi GTX-680 but my solution might be helpful.

When I first got my gigabyte gtx-780 it was useless for DCS in particular, nvidia drivers crashing / resetting no matter what versions were used.

Most other high end games worked fine however.

So after doing some searching and monitoring of the cards voltages and load levels, I was finding that because dcs actually doesn't stress it to badly, it would use the lowest possible voltage of 0.8xx - 0.9xx and a lowly 600mhz clock frequency, these would spike to higher levels depending on the scene being drawn, but almost all of the time you went to the map view or the action abaited, these levels dropped to the lowest dynamic voltage and clock speeds, and kaboom the dreaded black screen problem, it would flicker between blank black and back to the dcs picture a few times, then would stall the video driver altogether, requiring a total computer shutdown/ restart (most annoying).

Long story short, the solution for me was to reprogram the bios on the graphics card so it never went below 1.075 volts ever, that was the sweet spot for mine (the default for the kepler chips on the gtx-780 was approx 0.825v - 0.875).

So now the card runs no problems at all:

1.075v - 1.2v and 600 - 900mhz clock for old games that don't load up the graphics card to much, like old directx 9 titles, DCS, and general desktop work.

and

1.150 - 1.2v and 900 - 1071mhz clock for newer games that fully load the graphics processor, such as all Directx11 titles.

Unfortunately bios programming is a risky business, not for the faint hearted, but I couldn't get a reasonable response from Gigabyte them selves, even when I told them the solution, so I searched the net for the solution and programming tools and it took me 3 bios flashes to get it right for my purposes ;-).

Used:

Keplar Bios Tweaker v1.27 - to change voltage info of bios

Tech Power GPU-Z v0.7.8- for monitoring graphics card

Nvidia nvflash v5.142 - To actually write bios to graphics card.

Info and Bios Materials

http://www.overclock.net/t/1393791/official-nvidia-gtx-780-owners-club

for your sake

http://www.overclock.net/t/1232473/official-nvidia-gtx-680-owners-club

Hope this helps, as I know a lot of the nvidia models have this problem, it is the minimum voltage they use when the graphics engine is not loaded up and not the temperatures that are causing most of these issues. (This is my opinion)

Cheers, Ian

Try to limit max fps in graphics.lua - it can help. This helps me.

Hi,

To update you on what I've tried in the mean time: I went back to driver version 331.82 and set pci express setting to Gen 2 (it was indeed in auto and I'm running windows 8.1). Both did not help.

About the voltage, I can easily change that for my card using MSI afterburner (using it already to monitor the card and set the fan speed).

I didn't dare touching the voltage yet, is there a change I ruin my card playing with that?

Thanks for all your tips,

Marcel

New update: tried also to lock my fps (set it to 30). It didn't help (although it reduced the GPU temperature with about 10 degrees)

I had a look at the core voltage and right before the driver crash both the voltage limit and the OC voltage limit were triggered. Note I have not overclocked the card at all. What does this exactly mean and should I change something to the GPU voltage or does it mean that something is wrong with the card?

Regards, Marcel

Try reseating the video card in the motherboard...and check power connections too.

The video card could be on its deathbed though....time for an upgrade?

Best of luck!!

To give you an update I managed to get rid of the frequent resets of the GPU. I 'underclocked' the videcard until the problem went away. Is it normal that one must 'underclock' a videocard ot get it stable (at least for DCS)?

The "Voltage Limit" and "Overclock voltage limit" are still triggered though. Is this normal and what does this mean?

Regards, Marcel

To give you an update I managed to get rid of the frequent resets of the GPU. I 'underclocked' the videcard until the problem went away.

 

Is it normal that one must 'underclock' a videocard ot get it stable (at least for DCS)?

The "Voltage Limit" and "Overclock voltage limit" are still triggered though. Is this normal and what does this mean?

 

Regards, Marcel

Update on my previous post.

By lowering your MHZ or voltage, you are lowering the Wattage requirements from you PSU, which may be the case of the problem.

My best guess (theory) is that the video driver is trying to boost the clock MHZ higher than the default 3d mode is actually programmed for (maybe too high at a spurious rate - my highest boost clock speed is set at 1071 for a factory overclock, but I have caught it going higher than that, bios has the boost table at 1389mhz as its highest boost clock possible).

So my own problem seemed that the low end was under volted, and the upper end was over speeded, Changed the boost table highest 1389mhz to 1071mhz and made video card rock solid.

or

Your Power settings for a gtx-780 (example) are set to a certain maximum Wattage (225-300w, 250w std) max, if you try to exceed that wattage by running the GPU faster, one of Two things will happen the driver will throttle down the GPU to reduce the wattage load or the power supply will fail in some way, not being able to deliver that wattage (video display corruption, reboot, blank screen problem).

So the limits your seeing are that you can't draw anymore more wattage / power, as PSU is not able to supply it, so the driver is keeping it within operating limits.

So in summary:

• Have a 600w minimum PSU - that has the extra 2 x (2x4pin) Pci-e power connectors, and not the 2 x (2x3 pin) - this is what limits your wattage available.
  
• M/b with PCI-x 3.0 Slot
  
• Nvidia Drivers can vary - some work better than others
  

Side note

• I run a 620w corsair psu, with only 2x3pin Pci-e connectors with 2x4pin adapters, so I am limited to (75+75+75w=225watts), as the card needs 250watts, I should expect it to be throttled or have issues.
  
• I have tried the GPU with a seperate 620w psu with 2x4pin pci-e connects just running the GPU card (75w+150w+150w=375w) but card still played up the same as before I reprogrammed it, so was at factory defaults.
  

PCI-x Power specs

http://www.pcisig.com/developers/main/training_material...