Freezes then restarts nvlddmkm.sys

I just removed my riser and mounted my GPU stock standard and it seems to be running a lot better.
I can run a benchmark on 3Dmark without it crashing or an error popping up.
Check your mounting - try the second PCIE slot if you can, see if there is any difference.

Jebus said:

I am getting the same issue - I have a brand new 1000W PSU, just installed a new B550a-pro and running a 3090.
PC will just randomly hitch up and it might be ok, or it does a full restart.
I thought it was the PCIe gpu slot on my old Mobo (B550F GAMING WIFI) that was busted, as the problem went away when I used the 2nd PCI slot on that board. But I wanted to use the 1st slot to not get bottle-necked.

@jds63 Just wondering, have you mounted your GPU vertically with a riser?

Click to expand...

Hi and Welcome to the Forum !!

I also recently as you might of read purchased a new PSU thought maybe my old one was causing issue.
No my card is not vertically it’s horizontal. Never changed the slot but remounted enough times checking things inside PC.

Mentioning two different boards in your troubleshooting when switching PCIe slots. Saying problem went away when using second slot. What about on newer board ?
Don’t think it’s the Slot on board it’s rarely ever the thing to try for this issue in research a bad slot.
Thank for your input, still trying to figure as it can go month without it, then do it again…so on..

Jebus said:

I just removed my riser and mounted my GPU stock standard and it seems to be running a lot better.
I can run a benchmark on 3Dmark without it crashing or an error popping up.
Check your mounting - try the second PCIE slot if you can, see if there is any difference.

Click to expand...

As said before in my posts here mine runs fine gaming and stressing card with no issues at all, it randomly does it at times when i am just doing browsing or other lite work on PC. Just would freeze then restart or i would have to restart it cause it froze. Windows has a setting in system that make PC restart when issues occur method #5 here tell you [Solved] Dump File Creation Failed During Dump Creation
Get this issue also (volgmr) creates a memory dump.

I have been checking Nvidia's resource site for any clue as to what may be causing your issue and remember a similar issue with an older machine of mine fitted with an RTX2060 Super, this edit helped in my case and might in yours, it is worth a try.
This involves editing the registry & adding a key. Write down/ screen shot the location and the edit you made so you can undo it if needed.

1. Goto registry editor (start -> run -> regedit)
2. Navigate yourself to HKEY_LOCAL_MACHINE/System/CurrentControlSet/Control/GraphicsDrivers (make sure GraphicsDrivers is highlighted).
3. Right click -> New -> DWORD (32-bit) value -> Name it: TdrDelay -> double click it -> Set it to A (hexadecimal) or 10 (decimal).

When I used this key the only difference I saw was that the TDR issues stopped, your mileage may vary, it depends on why the issue occurs for you. I beleive the default value for this setting is 5 so all this edit does is to double it, possibly giving the driver extra time to respond to the calling function that is generating the error. If that makes sense.

DigitalGoat said:

I have been checking Nvidia's resource site for any clue as to what may be causing your issue and remember a similar issue with an older machine of mine fitted with an RTX2060 Super, this edit helped in my case and might in yours, it is worth a try.
This involves editing the registry & adding a key. Write down/ screen shot the location and the edit you made so you can undo it if needed.

1. Goto registry editor (start -> run -> regedit)
2. Navigate yourself to HKEY_LOCAL_MACHINE/System/CurrentControlSet/Control/GraphicsDrivers (make sure GraphicsDrivers is highlighted).
3. Right click -> New -> DWORD (32-bit) value -> Name it: TdrDelay -> double click it -> Set it to A (hexadecimal) or 10 (decimal).

When I used this key the only difference I saw was that the TDR issues stopped, your mileage may vary, it depends on why the issue occurs for you. I beleive the default value for this setting is 5 so all this edit does is to double it, possibly giving the driver extra time to respond to the calling function that is generating the error. If that makes sense.

Click to expand...

Thanks not sure if i read about this before.
Looks like i am setting the response time for a TDR to occur when a fault or error occurs.
Testing and debugging TDR during driver development - Windows drivers
See lot's of articles on this now.

One time running AIDA64 stressing GPU trying to produce issue maybe it gave me a warning about my TDR being low and may cause black screen during stress test, but it did not anyway.

Hardware acceleration is something else i heard about turning off for this issue. Found in Settings>Display>Graphics>Default Graphics Settings.
Remember this issue does not happen consistently enough to know right away.

Do you feel this might be a hardware fault from reading through this thread here ? If so changing some regedit can that have an effect on it more ?
For one i do not feel it is hardware, like another opinion.

​

Mine does not show TdrDelay so some article i read is to create it, but decimal of 60 not hexidecimal.
GPU drivers crash with long computations (TDR crash) - Substance 3D Painter

jds63 said:

Thanks not sure if i read about this before.
Looks like i am setting the response time for a TDR to occur when a fault or error occurs, does 5 mean seconds. ?

One time running AIDA64 stressing GPU trying to produce issue maybe it gave me a warning about my TDR being low and may cause black screen during stress test, but it did not anyway.

Hardware acceleration is something else i heard about turning off for this issue. Found in Settings>Display>Graphics>Default Graphics Settings.
Remember this issue does not happen consistently enough to know right away.

Do you feel this might be a hardware fault from reading through this thread here ? If so changing some regedit can that have an effect on it more ?
For one i do not feel it is hardware, like another opinion.

​

Click to expand...

It is so hard to tell really if an error is hardware or software related, especially an intermittent issue.
I think the value in that key is milliseconds, 5 seconds would be a long time in modern computer terms.

Turning off HAGs in Windows settings is certainly a valid troubleshooting option as is turning off resizable bar in the BIOS and any overclocks of CPU, RAM and GPU (so basically run everything at stock settings).

If the BIOS allows setting the NVME and PCIe speeds then changing them to the speed supported by the fitted hardware is also a good idea, so if using a Gen3 GPU on a Gen4 slot set the slot to Gen3 manually, sometimes an Auto setting can get it wrong.
Worth mentioning here to turn off any BIOS option for Fast Startup while troubleshooting.
If you are using Windows Core Isolation then turning that off can also help while testing.

If you are thinking a hardware fault (specifically with the GPU) then removing as many non essential hardware items and software components/ settings as possible from the mix for troubleshooting is the minimum action to take.
I would use something like Unigenes Superposition benchmark tool (there is a free version) to test the card as it allows you to run custom settings including large textures which will test all available VRAM, which a lot of other benchmarks don't do. And don't just run a test once, run it for a couple of hours or more as you want the card to be at max operating temperature while being stressed.
Cinebench on the other hand is a CPU related benchmark/ stress tool or you could run Prime95 with small data sets on a loop to really hammer the CPU to test for errors and temperature stability.

DigitalGoat said:

Turning off HAGs in Windows settings is certainly a valid troubleshooting option as is turning off resizable bar in the BIOS and any overclocks of CPU, RAM and GPU (so basically run everything at stock settings).

Click to expand...

Everything is at stock settings

To add to my list, using MSI Afterburner and / or HWInfo64 (available as a portable version) to monitor and record temps, voltages and load on system components like RAM can also be extremely useful.

DigitalGoat said:

Turning off HAGs in Windows settings is certainly a valid troubleshooting option as is turning off resizable bar in the BIOS and any overclocks of CPU, RAM and GPU (so basically run everything at stock settings).

Click to expand...

Everything is at stock settings, no overclock

DigitalGoat said:

To add to my list, using MSI Afterburner and / or HWInfo64 (available as a portable version) to monitor and record temps, voltages and load on system components like RAM can also be extremely useful.

Click to expand...

Used AIDA64 while running tests temps were good always on all CPU,GPU

Read article on Registry i wrote ?

Like said i have stressed tested it through AIDA64, Furmark before and even Unegine with no crashes and temps were fine on card.
Even i my thread have tested RAM with Memtest and through Windows 11

DigitalGoat said:

If the BIOS allows setting the NVME and PCIe speeds then changing them to the speed supported by the fitted hardware is also a good idea, so if using a Gen3 GPU on a Gen4 slot set the slot to Gen3 manually, sometimes an Auto setting can get it wrong.
Worth mentioning here to turn off any BIOS option for Fast Startup while troubleshooting.
If you are using Windows Core Isolation then turning that off can also help while testing.

Click to expand...

Not sure setting for this new GPU is a Gen 4 and on a Gen 3 PCIe slot on this Motherboard, even though they say it's compatible.

DigitalGoat said:

If the BIOS allows setting the NVME and PCIe speeds then changing them to the speed supported by the fitted hardware is also a good idea, so if using a Gen3 GPU on a Gen4 slot set the slot to Gen3 manually, sometimes an Auto setting can get it wrong.
Worth mentioning here to turn off any BIOS option for Fast Startup while troubleshooting.
If you are using Windows Core Isolation then turning that off can also help while testing.

Click to expand...

Double checked in BIOS M.2 i have is set right remember asking about this here when i bought one. It has 3 setttings on M.2 configuration Auto, SATA and PCIE (mine is on this).
Graphics card PCH Configuration has settings Auto (mine is on),Gen 1, Gen 2, Gen 3 (this one u want to move it to ?) Saying might not be right on Auto. RTX-3060Ti is a Generation 4 card but compatible with Gen 3 boards. Wouldn't speeds be slower on card ?

Setting M.2/ NVME and GPU Gen speeds to the closest match is ok, so a Gen4 GPU on a Gen3 slot is fine, PCIe is fine for NVME drives.
Speeds for GPU will not be affected as even a Gen4 card doesn't saturate a Gen3 bus.
Furmark is a power virus type tool, it is designed to run the GPU at max speed and temps/ voltages for extended periods to test for stability, unless it has been changed recently I don't think it uses all the available VRAM, nor does AIDA as far as I am aware.

Unigenes Superposition allows for testing all the cards VRAM which should show up any faults in the upper range of memory that may only be accessed under specific loads. Although you don't seem to be having the issue occur during high load situations anyway, but for peace of mind as far as hardware is concerned it might be worth testing.
Memtest in Windows 11 is not really sufficient, you should try memtest86 (this might have had a recent update, worth checking), a self booting memory tester that works outside the Windows environment is always better, I realise you have already tested the memory previously but it doesn't hurt to test it under different configurations.

I really do not think this is a hardware issue, I think, like you initially said, that this is a driver/ resource conflict or maybe even an error in coding for a driver, service or app, so I would concentrate troubleshooting in this area if all the hardware passes some basic but hefty testing.
One thing that came to mind is that you mention fitting a soundcard, maybe the issue is actually to do with a conflict between the soundcard drivers and the GPU's HDMI audio funtions, this theory might be a stretch though.

Jebus said:

I am getting the same issue - I have a brand new 1000W PSU, just installed a new B550a-pro and running a 3090.
PC will just randomly hitch up and it might be ok, or it does a full restart.
I thought it was the PCIe gpu slot on my old Mobo (B550F GAMING WIFI) that was busted, as the problem went away when I used the 2nd PCI slot on that board. But I wanted to use the 1st slot to not get bottle-necked.

@jds63 Just wondering, have you mounted your GPU vertically with a riser?

Click to expand...

"VIDEO_TDR_FAILURE" is most likely poor PCI-E contact/card "disappeared" from Windows.

DigitalGoat said:

Unigenes Superposition

Click to expand...

Try this one when I can. Still feel some of stress tests from aida64 engineer, unigenes heaven, furmark and playing new Pc games for hours pretty suffice t to say graphics card in most likely not issue or would have crashed then.

DigitalGoat said:

Although you don't seem to be having the issue occur during high load situations anyway

Click to expand...

Exactly

DigitalGoat said:

conflict between the soundcard drivers and the GPU's HDMI audio funtions

Click to expand...

Using display port. Thought of driver issue but last sound card was sound blaster.

Take a look at my responses to the delay in registry because was different than u said. Thanks so far as you see not many people have responded to this thread. Maybe there just as stumped as me have no input to think of

I looked at your registry key for graphics, you need to create the new value in the right hand side so you would end up with five entries, the default value of 5 I mentioned is hardwired into the drivers so doesn't show an entry by default.
Just another thing I came across was similar driver resets/ crashes caused by something called the MPO layer, there are instructions on Nvidia's site on how to disable it to test, although you may loose some functionality (not too clear here on what exactly), might be worth looking into though.

What power management settings are you using? I use high performance... as the problem nvlddmkm.sys detailed by you, has also occurred on mine.

High performance mode fixed it.

PenPusher said:

What power management settings are you using? I use high performance... as the problem nvlddmkm.sys detailed by you, has also occurred on mine.

High performance mode fixed it.

Click to expand...

Thanks i read about that and do use and have used high performance