Witcher 3 crashes every 5-10 minutes, GPU disconnects, only happens in Witcher 3

+
Hey everyone, I need help with a frustrating issue that's been driving me crazy.
The Problem: Witcher 3 crashes between 5 to 10 minutes of playing. It feels like a GPU disconnect, the screen goes black and I have to fully power off and restart my PC. What makes this confusing is that every other game runs perfectly fine in longer sessions.
Games that run without any issues:
  • Skyrim SE with 400+ mods
  • Hogwarts Legacy
  • Baldur's Gate 3
  • Gotham Knights
  • League of Legends
  • DreadOut 2
My Specs:
  • CPU: i7 4790K (OC off)
  • RAM: 24GB DDR3
  • GPU: XFX RX 590 8GB
  • OS: Windows 10 Build 1904
  • Storage: 254GB SSD (Witcher 3 installed here), 500GB HDD, 254GB SSD
Note: Every time I install a driver I use DDU to fully clean before reinstalling.
Everything I've tried:
  • Verify game files
  • Driver updates and downgrade to 22.11.2 (surprisingly gave more FPS but still crashes)
  • Disabled cloud saves in both Steam and in-game
  • GPU power draw +15 in AMD Adrenalin
  • Disabled all overlays
  • Capped and uncapped FPS
  • Borderless and fullscreen
  • Lowered in-game settings as much as possible
  • Tried DX12 (way lower FPS and still crashes)
  • Disabled MPO in Regedit
  • Applied DXVK and dxgi in the game's exe folder (this helped slightly, crashes pushed to around 15 minutes but still happens)
  • Launched directly from the launcher
  • Tried a borrowed PSU from a friend (good quality) — still crashes
  • Temporarily swapped to a different motherboard — still crashes
  • Tried a newer save file
  • Even tried Witcher 3 Classic — still crashes
What's making this extra frustrating: I played Witcher 1 and 2 specifically to understand the story before jumping into 3, and now this is happening right at the finish line.
If anyone has experienced this or knows something I haven't tried, please help. Happy to provide any additional info needed.
 
For:
Temporarily swapped to a different motherboard — still crashes
Does that mean you swapped the HDD/SSD to a completely different PC?

If so, then I imagine what you have is going to be related to either something on the master file table that's interfering on that physical drive (corrupted .dll somewhere, bad sectors, etc,) or it's corruption of some sort with drivers.

I'm jumping the gun a bit, though. First, let's confirm my question above.
 
For:

Does that mean you swapped the HDD/SSD to a completely different PC?

If so, then I imagine what you have is going to be related to either something on the master file table that's interfering on that physical drive (corrupted .dll somewhere, bad sectors, etc,) or it's corruption of some sort with drivers.

I'm jumping the gun a bit, though. First, let's confirm my question above.
yes but just to check if the motherboard changing it fixes , not to mention it's the same as my motherboard , just newer one , mine is used
 
yes but just to check if the motherboard changing it fixes , not to mention it's the same as my motherboard , just newer one , mine is used
Okay, good -- that also narrows the problem down to most likely something on or with the drive itself. Is it an HDD or SSD? How old?

FIrst and easiest step:
Open a Command Prompt with Administrative Access (R-click on the icon and "Run as administrator.") Then, type the following at C:\__

sfc /scannow

(Note the space after "sfc.") Don't interrupt that scan for any reason. It will scan the entire file table and try to fix any issues it encounters. If it encounters problems, it may look like the process has hung, but it's still running. There's just no visual indicator. Normally, it takes 5-15 minutes or so, but it can seem to lock up for hours if there are any major issues.

When it's finished, it will create a log you can read for specifics.
 
Last edited:
Okay, good -- that also narrows the problem down to most likely something on or with the drive itself. Is it an HDD or SSD? How old?

FIrst and easiest step:
Open a Command Prompt with Administrative Access (R-click on the icon and "Run as administrator.") Then, type the following at C:\__

sfc scannow

(Note the space after "sfc.") Don't interrupt that scan for any reason. It will scan the entire file table and try to fix any issues it encounters. If it encounters problems, it may look like the process has hung, but it's still running. There's just no visual indicator. Normally, it takes 5-15 minutes or so, but it can seem to lock up for hours if there are any major issues.

When it's finished, it will create a log you can read for specifics.
"Microsoft Windows [Version 10.0.19041.264] (c) 2020 Microsoft Corporation. All rights reserved. C:\WINDOWS\system32>sfc /scannow Beginning system scan. This process will take some time. Beginning verification phase of system scan. Verification 100% complete. Windows Resource Protection did not find any integrity violations. C:\WINDOWS\system32>"


it only took like 2 minutes to complete , and i repeat i can play other game just fine, and i am using ssd and witcher 3 also installed on a ssd
 
/scannow -- I knew that looked funny. TY!

Well, good! If it that came back clean, then your hardware is likely in good shape.

That leaves us with data.

Are you using any mods, or have you ever used any in the past? Even if you removed all mods, that can still leave reference data in your save state that results in all ranges of weirdness, including crashes. Often removing mods is what causes it.

Also note, that reinstallation of the game will not correct such an issue if it occurs. It's something that exists in the save file forever more.

Testing for that is simple:
1.) Go to your Wicther 3 save file location:
C:\Users\[YOUR_USERNAME]\Documents\The Witcher 3\gamesaves\

2.) Cut and paste all of those files/folders to a backup location on the desktop, flash drive, etc. (You can just put them back when finished testing.)

3.) Go completely offline. (ideally, disconnect from your modem/router.) This will prevent the game from restoring saves from the cloud.

4.) Begin a brand new game and run around White Orchard for a bit. If the game doesn't crash, we know that the issue lies somewhere in the save files.

If, however, it does crash, we will know that it's something in the installation or system config (drivers, APIs, firmware, etc.) that aren't playing nicely together.

It's worthwhile testing even if it isn't mod-related, as there might be save data corruption for other reasons (borked update, save corruption from something like power loss, etc.)
 
/scannow -- I knew that looked funny. TY!

Well, good! If it that came back clean, then your hardware is likely in good shape.

That leaves us with data.

Are you using any mods, or have you ever used any in the past? Even if you removed all mods, that can still leave reference data in your save state that results in all ranges of weirdness, including crashes. Often removing mods is what causes it.

Also note, that reinstallation of the game will not correct such an issue if it occurs. It's something that exists in the save file forever more.

Testing for that is simple:
1.) Go to your Wicther 3 save file location:
C:\Users\[YOUR_USERNAME]\Documents\The Witcher 3\gamesaves\

2.) Cut and paste all of those files/folders to a backup location on the desktop, flash drive, etc. (You can just put them back when finished testing.)

3.) Go completely offline. (ideally, disconnect from your modem/router.) This will prevent the game from restoring saves from the cloud.

4.) Begin a brand new game and run around White Orchard for a bit. If the game doesn't crash, we know that the issue lies somewhere in the save files.

If, however, it does crash, we will know that it's something in the installation or system config (drivers, APIs, firmware, etc.) that aren't playing nicely together.

It's worthwhile testing even if it isn't mod-related, as there might be save data corruption for other reasons (borked update, save corruption from something like power loss, etc.)
I think I already mentioned the stuff I tried, like a fully fresh save file and no mods at all. I reinstalled the game four times—two times for the classic version. To begin with, cloud saves are off in the game (cross-save), and also turned off in Steam, and I tried playing in offline mode as well.


It’s really hard to tell what’s causing it because other games are working fine. But right now, my GPU has a switchable BIOS, like it has a stealth mode and a performance (overclock) BIOS. All of the crashes happen on the stealth BIOS. I completely forgot about the performance BIOS—I have yet to try that. That’s probably the one thing I still need to test.
 
I think I already mentioned the stuff I tried, like a fully fresh save file and no mods at all. I reinstalled the game four times—two times for the classic version. To begin with, cloud saves are off in the game (cross-save), and also turned off in Steam, and I tried playing in offline mode as well.


It’s really hard to tell what’s causing it because other games are working fine. But right now, my GPU has a switchable BIOS, like it has a stealth mode and a performance (overclock) BIOS. All of the crashes happen on the stealth BIOS. I completely forgot about the performance BIOS—I have yet to try that. That’s probably the one thing I still need to test.
TW3's GPU demands can go from 0 to 60 in just a few frames. Very possible that the stealth mode could be causing it. That sort of thing is normally power-saving, so if it's not providing enough voltage to the card, that could result in a full-system hang. Easy test, again. Good thinking!

As for the modding / save file issue, it's extremely particular. Starting a new game but saving over an existing game could mean that the bad reference data is not overwritten, introducing the issue into that playthrough. This is true of quicksave slots as well. So, starting a new game with existing quicksaves would instantaneously copy the bad reference data over during the first autosave in the tutorial. Similarly, loading an existing game and saving in a new slot would simply copy the bad reference data over, creating the same situation.

Reinstallation also wouldn't matter for mod / save file issues. Such problems are in the save data, not the game files.

Only by beginning a brand new vanilla game with a completely empty save folder would it create a 100% clean save. If you have not done exactly that give it a try. Only takes a minute to empty the folder and put everything back when you're done.

If you did try exactly that, bummer. Let's hope it's a stealth mode voltage or overheating thing.
 
Last edited:
TW3's GPU demands can go from 0 to 60 in just a few frames. Very possible that the stealth mode could be causing it. That sort of thing is normally power-saving, so if it's not providing enough voltage to the card, that could result in a full-system hang. Easy test, again. Good thinking!

As for the modding / save file issue, it's extremely particular. Starting a new game but saving over an existing game could mean that the bad reference data is not overwritten, introducing the issue into that playthrough. This is true of quicksave slots as well. So, starting a new game with existing quicksaves would instantaneously copy the bad reference data over during the first autosave in the tutorial. Similarly, loading an existing game and saving in a new slot would simply copy the bad reference data over, creating the same situation.

Reinstallation also wouldn't matter for mod / save file issues. Such problems are in the save data, not the game files.

Only by beginning a brand new vanilla game with a completely empty save folder would it create a 100% clean save. If you have not done exactly that give it a try. Only takes a minute to empty the folder and put everything back when you're done.

If you did try exactly that, bummer. Let's hope it's a stealth mode voltage or overheating thing.
Performance BIOS did not fix it. I'm going to give up for now. Maybe one day I'll play it, but unfortunately not with this current setup. Thanks for taking the time to reply
 
Performance BIOS did not fix it. I'm going to give up for now. Maybe one day I'll play it, but unfortunately not with this current setup. Thanks for taking the time to reply
Sorry it's being problematic. If you ever feel like fiddling some more, we'll be here!
 
Sorry it's being problematic. If you ever feel like fiddling some more, we'll be here!
Hey, I'm happy to let you know that reducing the maximum GPU clock speed by 100 MHz, lowering the power draw by 20%, and cranking the fan settings to the maximum in AMD Adrenalin (Gaming → The Witcher 3 → Tuning) fixed the problem.

Now, the GPU temperature has dropped from around 85°C to 70°C, and the power draw has decreased from 180W to 150W. Even though my PSU and GPU should be able to handle the higher temperatures and power draw, other games run perfectly fine at 85°C and 180W without any issues.

Of course, I'm getting lower FPS than expected, but at least it's no longer crashing after 5 minutes!

I hope this helps someone in the future who runs into the same issue.
 
Top Bottom