GPU Crashing System From Hibernation

[Deadmano]

For 2 years now I have been experiencing this issue under Windows 11. Initially I thought it was my RTX 3080 going bad, but ever since I've gotten my RTX 5090 this issue still persists!

If I hibernate my PC and at a later stage resume it, and try to launch any 3D application (game), there is a high chance my entire PC will freeze and require a complete hard reset, after which it is fine.

The only thing that shows up in event viewer prior to the crash is:

The description for Event ID 153 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer.

If the event originated on another computer, the display information had to be saved with the event.

The following information was included with the event:

\Device\Video3

Error occurred on GPUID: 100

The message resource is present but the message was not found in the message table

I have tried reinstalling countless versions of different NVIDIA drivers, always using DDU to install cleanly, and this issue has never gone away. Windows 11 is up to date, so are bios and hardware drivers.

If anyone has any ideas on what I could try, I'd appreciate it! The error itself is rather strange, I've tried the "dll" suggestions to get it to properly register for event viewer, but that did not help.

 

[K7SEM]

Need to know about complete specification in order to narrow down the problem
First guess is the PSU need to replaced with a higher wattage and good quality
Second guess is the whole computer components temperature running too hot
Third guess is the join of screw and board a bit loose

 

[nomdeplume]

If anyone has any ideas on what I could try, I'd appreciate it! The error itself is rather strange, I've tried the "dll" suggestions to get it to properly register for event viewer, but that did not help.

This is interaction between hardware and OS. Somewhere a setting or firmware is at fault. Simple means of determination is fresh install on unused drive with minimal installs required to test reoccurrence of issue. Something you can test a wide range of settings on.

Can't fully remove PSU or mobo (or CPU) are showing signs of degradation. More likely to be software than hardware if there is no issues going from idling to hard use and back to idle.

 

[Deadmano]

Need to know about complete specification in order to narrow down the problem
First guess is the PSU need to replaced with a higher wattage and good quality
Second guess is the whole computer components temperature running too hot
Third guess is the join of screw and board a bit loose

Full system is here.

I doubt it is the PSU, as I had a Corsair AX860i PSU prior, and now a Asus ROG Thor 1200P 80+ Platinum PSU. Both are high-end with more than enough wattage, the RTX 3080 using the AX860i and the RTX 5090 using the 1200w Thor now.

Temperatures are low throughout, confirmed with HWiNFO, power on the 12v rail never dips below 12.3v. Voltages throughout look nominal, CPU never peaks above 75c during gaming (far from intended operational limits) and the GPU at a cool 65c. So definitely not temperature related. Solid Phanteks NV5 chassis with 12 fans, incredible airflow.

A screw on the motherboard causing a short? Unlikely. The PC is rock-solid through gaming or benchmarks; this issue only occurs after resuming from hibernation.

 

[Deadmano]

This is interaction between hardware and OS. Somewhere a setting or firmware is at fault. Simple means of determination is fresh install on unused drive with minimal installs required to test reoccurrence of issue. Something you can test a wide range of settings on.

So I've done 3 clean OS installs between the period of having the RTX 3080 and now the RTX 5090. It almost feels like it is GPU driver related (hence why I posted in the GPU forum section) and some searches have people point out the DP cable, or TDR length, or G-Sync as potential issues. I've changed DP cables and the issue persists. My GPU only has 1 HDMI port and I have 3 monitors, else I'd have tried the HDMI route but that seems overkill.

Can't fully remove PSU or mobo (or CPU) are showing signs of degradation. More likely to be software than hardware if there is no issues going from idling to hard use and back to idle.

So thankfully all 3 were swapped out in a recent build (upgraded) and I still get the same issue. Which is what makes me feel like it is driver-related. Of course I'd be happy to try out other things, but I just find it strange that through pretty much a new rig the issue persists, and only with resuming from hibernation... There's also a strange case where a recent VR game (Half-Life Alyx) causes the same system-wide PC freeze) noticed one CPU core spiked at 100% prior), but this might be an unrelated issue as I've seen reports of others complaining HL:A freezes their entire machine.

 

[nomdeplume]

The PC is rock-solid through gaming or benchmarks; this issue only occurs after resuming from hibernation.

My active use and awareness of command line / registry tinkering that might aid resolving where the conflict resides has fallen off. Switching hibernation off and deleting the file might force a refresh when you set it back up.

Can I ask why you are using hibernation and what size you have it set up to use?

 

[Deadmano]

Can I ask why you are using hibernation and what size you have it set up to use?

A ton of multi-tasking for work, so much easier to resume the next day without having to re-open all the various tooling needed.

The size would be the default for Windows 11, so 40% of the system RAM, in my case ~25.6GB.

It's a very strange issue. I can go days without a problem, but there will be one time I resume from hibernate and if I launch a demanding game the system will just straight up freeze.

Note that the system doesn't immediately freeze upon resuming from hibernation; I could work on it (no gaming) for hours without issue, hibernate again, resume the next day, etc.

Today I did some browsing for a few hours after resuming from hibernation, and then I launched Skyrim Special Edition (a game I haven't touched in years) and then immediately got the hard freeze. Not a BSoD. Everything is up, but frozen. My keyboard has RGB lighting and I notice after about ~30 seconds in this "frozen" state the LEDs turn off, indicating the Corsair service connection must have been interrupted. That's when I know there is no recovering and need to force restart by holding the power button. Check event viewer and the only log in there is the one in the quote in my original message "The description for Event ID 153 from source nvlddmkm cannot be found" which makes me believe there'd be an actual reason but for some reason the system can't share it because the installation is "corrupted"? DDU'd the NVIDIA drivers (full uninstall/clean) and installed new ones; never solves it. SFC/DISM does not fix it either. And a reinstall of Windows did not either. :/

 

[nomdeplume]

It's a very strange issue. I can go days without a problem, but there will be one time I resume from hibernate and if I launch a demanding game the system will just straight up freeze.

So game drive is potentially the issue. Try setting OS power plan to never let drives sleep and see if problem goes away. Wouldn't hurt to see if there is a firmware update for that drive either.

Surest way to eliminate GPU driver is DDU and install one from before the issue started.

 

[Visible Noise]

A ton of multi-tasking for work, so much easier to resume the next day without having to re-open all the various tooling needed.

How to enable restartable applications in Windows 11

The ability to restart applications when booting into Windows 11 is built in. Here’s how to enable it for registered third-party applications.

www.techrepublic.com

 

[Deadmano]

So game drive is potentially the issue. Try setting OS power plan to never let drives sleep and see if problem goes away. Wouldn't hurt to see if there is a firmware update for that drive either.

Sadly I use NVMEs for all my game drives. So no sleeping drives here! In fact there's only a single HDD (replaced two mechanicals recently for SSDs) and that is set to never sleep as well.

Surest way to eliminate GPU driver is DDU and install one from before the issue started.

In my very first post I mentioned the steps I took with trying various NVIDIA drivers. And DDU was there for every step of the way. Good call though!

How to enable restartable applications in Windows 11

The ability to restart applications when booting into Windows 11 is built in. Here’s how to enable it for registered third-party applications.

www.techrepublic.com

Thanks, though this brings a lot of side effects as I may not always want certain things open all the time. Restarting means I get a clean boot with the bare minimum running. Good for certain use-cases, but not for mine where things change all the time depending on the time of project I am working on at the time (we use various tooling on a per-project basis). I appreciate you sharing it though!

 

[Deadmano]

Wouldn't hurt to see if there is a firmware update for that drive either.

And I double-checked but Kingston SSD Manager and Western Digital Dashboard have no firmware updates for any of my SSDs or the one HDD.

 

[Veseleil]

A ton of multi-tasking for work, so much easier to resume the next day without having to re-open all the various tooling needed.

Hibernation is not your friend. It introduces a lot of potential problems, as it needs to save all the data onto a drive, and then restore the state after.
It's a laptop feature that sucks on a desktop, especially win 11, and it always was a bad idea IMO.
Turn the PC off when it's not in use. Learn to use startup folder for launching apps that you think you need right after an OS starts.
Problem solved.

 

[tugrul_LordOfDrinks]

Enable screensaver, make sure cpu goes in idle state too, maybe also enable pcie power savings, pc will be good. Hybernation is not a must.

 

[_roman_]

How old is the firmware of graphic card / mainboard?

I would check the mainbaord also:
My asus mainboard has uefi issues with certain settings to crash my intel wlan module. Crashed once a week, but that happend 100%. Some power saving feature fault of asus mainboard or fault of intel wlan module. Both sold junk hardware (problematic) in past few years.

 

[R0H1T]

It's a laptop feature that sucks on a desktop, especially win 11, and it always was a bad idea IMO.

Well I've been using it since win7 & it's never been bad, unless the PC/OS itself was unstable somehow! Could be display driver timeout issue, check what's the value here ~

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers
TdrDelay=
TdrDdiDelay=

 

[evernessince]

Not a solution to fix hibernation but Atomware ToolBox allows you to save PC State and restore.

 

[Deadmano]

Enable screensaver, make sure cpu goes in idle state too, maybe also enable pcie power savings, pc will be good. Hybernation is not a must.

Be mindful that not everyone is as fortunate to be able to leave their PC running 24/7. I stay in a country where power cycles (known as loadshedding) are frequent, and we get given allocated hours in which we have electricity which alternates with times where we do not have electricity. In my case, for my own sanity, being able to resume from where I left off abruptly due to these constraints is a must.

How old is the firmware of graphic card / mainboard?

The graphics card is a few months old at best (the RTX 5090 just came out) there are no firmware updates for my model. And the motherboard is on the latest BIOS which came out ~a month ago.

Well I've been using it since win7 & it's never been bad, unless the PC/OS itself was unstable somehow! Could be display driver timeout issue, check what's the value here ~

There are no keys with those names.

Just to reiterate as I feel this might get lost through the responses;

My PC is stable under all conditions (synthetic benchmarks/gaming/intense 3D rendering in Blender) unless I hibernate and resume at a later stage; in which case on the next 3D application (generally a game) my PC will lock up and freeze. If I resume from hibernation and just use the PC normally all is fine, and I can do this for upwards of a week without any issues. There seems to be an issue with initialising the 3D drivers, as per the event viewer log I shared in my initial post. DDU'd across 10 different drivers has made no difference, and this issue has persisted across 2 different GPUs (RTX 3080 and now an RTX 5090) and across an entire different rig swap of Motherboard/CPU/RAM/PSU. I don't believe it is hardware-related, but rather on the OS level (Windows 11, latest updates) and some incompatibility somewhere. I've checked to ensure I do not have any OSD-type applications/overlays that might conflict (RTSS/Discord etc).

While that normally is an inconvenience and I simply get around it by performing a clean restart if I want to do anything serious, I am more concerned about the recent case of a VR game (specifically VR, all other games are fine) causing the same system-wide freeze. Half-Life: Alyx apparently has done this for several other people, and I guess the combination of the hibernation resume and this has gotten me to wonder if there is an actual issue with my system, or rather two unrelated things, one that can be overcome by simply restarting the PC when possible chalking it up to just "Windows being Windows" (never freezes on a clean boot) and the other the game which sadly doesn't matter if it is a clean boot or not, will eventually cause the system to halt (no BSOD) regardless of clean boot or not which I believe might be a memory leak as the CPU usage throughout the game is at around ~40% never peaking above that, and when this issue occurs where the entire system freezes the CCD0 (gaming core on the R9 7950X3D) has all its processors at 100% usage, I know this because I was monitoring resource usage and when the system froze I saw all cores at 100% when prior to that they were not even close.

 

[R0H1T]

There are no keys with those names.

You need to create them then, New > DWORD (32bit) Value & enter values above 2 (TdrDelay) and 5 (TdrDdiDelay) which are defaults IIRC. I use 8 & 16 for them it has helped me a fair bit, doesn't solve every crash ever but the general instability is gone especially with graphics intensive programs or games.

 

[Deadmano]

You need to create them then, New > DWORD (32bit) Value & enter values above 2 (TdrDelay) and 5 (TdrDdiDelay) which are defaults IIRC. I use 8 & 16 for them it has helped me a fair bit, doesn't solve every crash ever but the general instability is gone especially with graphics intensive programs or games.

I remember this being a thing back in the days of Windows 7; is this still needed/helpful in 2025 under Windows 11? Do you have any tested sources I could read up on, or is this just from your own personal experience? What OS and GPU do you use? I don't mind giving it a try, I just don't want to add any further system instability with adjusting these values used to cause back on Windows 7.

 

[R0H1T]

I wouldn't say it's "needed" but if you are having similar issues it helps ~

https://helpx.adobe.com/substance-3d-painter/technical-support/technical-issues/gpu-issues/gpu-drivers-crash-with-long-computations-tdr-crash.html

It's helped me with (chrome) browser crashes as well, although it's probably also to do with their regular updates. If you search for TdrDelay or TdrDdiDelay you would see a lot more hits even post 2020 or so.

I'm on win11 with RTX 3080 (12GB) ~

 

[Deadmano]

Thanks @R0H1T, I'll give that a try and see if it helps, it's worth a shot anyway!

Still more concerned why event viewer says "The description for Event ID 153 from source nvlddmkm cannot be found".

It mentions the installation might be corrupted, but that's a part of the NVIDIA driver suite, so a full reinstall (via DDU) surely should have resolved that?

 

[R0H1T]

I generally also test stability over Linux, if you have the time try running a live distro & see if you can run some (graphics) tests from there. This should ideally rule out any hardware issues, if that's stable then there's some issues with Windows. If not an option (Linux) I'll add those registry options, then do a clean install of studio drivers, all at stock setting, & see if that works.

The event log thing doesn't really tell you too much unless you have the memory dump. If the system can't write to event log, for some odd reason, before windows restarts this could happen.

 

[Deadmano]

Just as a follow-up to the TDR suggestion, I came across this:

TDR stands for Timeout Detection and Recovery. This is a feature of the Windows operating system which detects response problems from a graphics card, and recovers to a functional desktop by resetting the card. If the operating system does not receive a response from a graphics card within a certain amount of time (default is 2 seconds), the operating system resets the graphics card.

Before TDR existed, problems of this nature would have resulted in a system freeze and required a reboot of the operating system. If TDR is enabled and you see the TDR error message, "Display driver stopped responding and has recovered," this means that the Windows operating system reset the display driver.

What's interesting is that I have never seen the error message for TDR, which led me to believe that TDR was never even working to begin with, since the symptom of not having TDR enabled means the system would just freeze and require a reboot... Curious, I dug a bit deeper and noticed there is a "TdrLevel" key, info available here. By default TdrLevel is meant to be on 3 (Recover on timeout (default value) yet mine was set to 0 (Detection disabled) which I have no idea why that would be the case since I have never touched any values on this system via the registry...

The linked article from Adobe suggests a TdrDelay and TdrDdiDelay of 60 each, so 60 seconds; any downside to setting the values that high compared to the default of 2 and 5?

I am hoping this might be the case, and allow my PC to at least recover from a GPU crash (if that is even the case here) and at worst just crash the game instead of my system! Fingers crossed!

 

[R0H1T]

Those values are too high IMO but YMWV, I went with lower values because they just worked for me. If it wasn't stable I could've gone higher. Those high values could work for specific programs which probably need the longer delay/timeout for whatever reason.

Didn't need to tinker with it so can't say much about that ~

By default TdrLevel is meant to be on 3 (Recover on timeout (default value) yet mine was set to 0 (Detection disabled) which I have no idea why that would be the case since I have never touched any values on this system via the registry

Yeah that's an issue with programs/games on Windows. Practically any of them can do that with admin access, during installation, this is why there's so much instability as compared to say Linux.

 

[Assimilator]

https://learn.microsoft.com/windows-hardware/drivers/display/tdr-registry-keys for full reference of the TDR registry keys.

Be mindful that not everyone is as fortunate to be able to leave their PC running 24/7. I stay in a country where power cycles (known as loadshedding) are frequent, and we get given allocated hours in which we have electricity which alternates with times where we do not have electricity. In my case, for my own sanity, being able to resume from where I left off abruptly due to these constraints is a must.

South Africa?

Still more concerned why event viewer says "The description for Event ID 153 from source nvlddmkm cannot be found".

It mentions the installation might be corrupted, but that's a part of the NVIDIA driver suite, so a full reinstall (via DDU) surely should have resolved that?

This log, in various flavours, appears on every computer with NVIDIA drivers. It means that the driver reported error code 153, but no specific message along with it. This is pretty normal - 153 seems to be NVIDIA's generic "something went wrong but we don't know quite what" code; there's a massive reddit thread about this particular code with no apparent rhyme or reason as to what fixes it.

The linked article from Adobe suggests a TdrDelay and TdrDdiDelay of 60 each, so 60 seconds; any downside to setting the values that high compared to the default of 2 and 5?

I'd start with setting TdrLevel to 3 and leaving the other values. If that doesn't help, double the defaults (so 2 => 4 and 5 => 10) and try again. Repeat until your system either works or you hit over a minute on either value; if the latter then no amount of delay is going to help.