nvlddkmkm and dxgkrnl drivers crashing after waking up system from sleep / monitor off

[Siansi]

Hi,

I have a weird issue and i am just a step away before I RMA my video card, i do not have any more ideas.

My Problem Everytime after the PC turns off the displays or goes intor sleep due to inactivity the displays will not wake up again. After inspecting the minidumps created from windows I found that the dxgkrnl.sys and nvlddmkm.sys drivers are crashing and preveting the gpu to turn on. The Issue ist not consistent but has a probability of  roughly 70% to occur. The GPU works fine in every Game I played and did not once crash otherwise, I own the card since January 2021. It also passes benchmarks with expected results.

Workaround Just power off the PC and turn it back on againg. A reset with the reset switch does not work.

 

Tried soulutions All the things I tried to fix the issue

• Clean install of the drivers with (currently back running with the Drivers from GeforceExperience not Windows Update)
  • DDU
  • the Nvida tool
  • Windows update
• Cean install of Windows
  • used booth Win10 and now Win11
• Problem occourd with Ryzen 2700X and now Intel i7 12700K and of course the accorinding differnt Motherboards
• I did turn off
  • Hibernation via the CMD
  • The energysaving state through windows powermanagament
  • PCI powersaving
  • GPU powersaving with the NVIDIA Controlpanel
  • (note: just turning off the displays after x time of inactivity seems to be enough to trigger the Issue)
• Switched from the OC VBIOS of the GPU to the Silent VBIOS
• Updated the Silen VBIOS from F10 to F12 Version

My Hardware

• Intel i7 12700K
• Gigabyte Z690 UD DDR4
• Gigabyte RTX 3070 Gaming OC 8GB Rev 1.0
• 32GB (4 x 8GB) G.Skill Trident Z RGB DDR4-3200 CL14
• 500GB 970 Samsung Evo M.2 (Boot Drive)
• 750W Seasonic Prime Ultra Modular 80plus Gold
• 1TB 660P M.2 and 4TB Seagate Barracude for Storage

Dump File

All the files contain very similiar information and the order of the shown eventy is the same.

Spoiler

 

Thank you for reading until here and thank you for answering in advance.

 

Regards Siansi

Edited September 25, 2022 by Siansi
Forgot the mention that the card works fine in every other scenario.

[Bombastinator]

Got another workaround for you but it will likely waste power:  shut off your monitor power saving and your sleep function completely.   You just have to shutdown whenever you leave the machine.  It’s the way computers used to work. Sleep states and monitor energy saving are fairly new things.  You may want to install a screensaver though in case you walk away and leave it on to prevent burn.  LCDs don’t have anything like the problems CRTs, plasmas, or OLEDs have, but I don’t know what you have.

[Siansi]

This is what I am currently doing. But in some edge cases like locking the PC because I have to afk very quick and I want to return anyway, the powersettings are somehow circumvented by windows. And it grinds my gears having to forcefully shutdown the pc everytime and beeing extra aware to prevent data loss for open work.

[Bombastinator]

5 hours ago, Siansi said:

This is what I am currently doing. But in some edge cases like locking the PC because I have to afk very quick and I want to return anyway, the powersettings are somehow circumvented by windows. And it grinds my gears having to forcefully shutdown the pc everytime and beeing extra aware to prevent data loss for open work.

Yep.  They’re auto.  And the defaults are moderately aggressive. You have to change them, and do it for all the settings.  I vaguely remember there are three in windows.  May also be some in bios.  Not sure.  It’s an old memory and bioses vary a lot.  Control panel -> power settings or something, could be other places too though.  Iirc there at least used to be one in the Nvidia junk.

[Robchil]

sleep states have always been unstable.. it depends on the stability of your memory, disk and how long the system has been running. like file fragmentation etc. 

to make anything more stabile unless it's a notebook where you use it actively i would turn off anything in bios regarding hibernation and sleep configuration and turn it off entirely in windows. 

as said, use a timer to turn the display off if your not using the computer. 

[Siansi]

1 hour ago, Robchil said:

sleep states have always been unstable.. it depends on the stability of your memory, disk and how long the system has been running. like file fragmentation etc. 

to make anything more stabile unless it's a notebook where you use it actively i would turn off anything in bios regarding hibernation and sleep configuration and turn it off entirely in windows. 

as said, use a timer to turn the display off if your not using the computer. 

Already did that and now figured that the problem also occurs when the display turns off and the PC does not go into standby/energy/hibernation. I have to set the setting for turn off display to never if I want to avoid the error. Like that:

 

I also disabled S2 Sleepstate and ErP in BIOS.


(Sorry its in German but you can guess the menu by the layout. Niemals = Never).

 

5 hours ago, Bombastinator said:

Yep.  They’re auto.  And the defaults are moderately aggressive. You have to change them, and do it for all the settings.  I vaguely remember there are three in windows.  May also be some in bios.  Not sure.  It’s an old memory and bioses vary a lot.  Control panel -> power settings or something, could be other places too though.  Iirc there at least used to be one in the Nvidia junk.

Yeah there is a "always choose best performance" or smth. like that. I did set that also.

Im just wondering because I did build some PCs for Friends and Familiy with very differnt hardware and never encounterd such a annoying problem. But maybe it could be hardware related? I can source a GTX 1050 and try it for a change or just RMA the Card too see if there is any defect. Luckly we have a mandatory 2year warranty in the EU Should it be software related yeah ... then I can just resort to workaround and hope for some fix by MS or NVIDIA.

 

[Bombastinator]

1 hour ago, Siansi said:

Already did that and now figured that the problem also occurs when the display turns off and the PC does not go into standby/energy/hibernation. I have to set the setting for turn off display to never if I want to avoid the error. Like that:

 

I also disabled S2 Sleepstate and ErP in BIOS.


(Sorry its in German but you can guess the menu by the layout. Niemals = Never).

 

Yeah there is a "always choose best performance" or smth. like that. I did set that also.

Im just wondering because I did build some PCs for Friends and Familiy with very differnt hardware and never encounterd such a annoying problem. But maybe it could be hardware related? I can source a GTX 1050 and try it for a change or just RMA the Card too see if there is any defect. Luckly we have a mandatory 2year warranty in the EU Should it be software related yeah ... then I can just resort to workaround and hope for some fix by MS or NVIDIA.

 

Interesting.  I thought it was 1 year.  I don’t live there though.

This 1050 is known good?  Might be a good idea.  even a 710 would work. Or an 8800gt or something. Power is irrelevant as long as it’s pcie. It’s always annoying to do an RMA, wait 6 weeks and have the thing come back with a note that says it’s fine in business English.  Or worse yet Chinese character and you have to get someone to translate it for you because OCR won’t pick it up.

Edited September 25, 2022 by Bombastinator

[Siansi]

I tested some more stuff with it and now I have sometimes half the right screen just coverd in black bars and glitched. I gonna RMA the card and post the results later here.

[Siansi]

Update: The GPU came back from the RMA and according to Gigabyte it was fine.

Ive tested around a little bit more and changed out my PSU for a 850W Seasonic Prime Gold and now everything works as should. My old PSU must have gone faulty.