Tried this, all it seemingly did is make the whole PC hang for longer before it gave the usual crash error. Sadly, making this solution a worse result than just accepting the crashes due to the added delay 
To put it plainly, this issue is halting production for our project and I can’t see a way forward.
Has anyone experienced these crashes with a github 5.2 / 5.3 engine build?
You also may look at this option.
You may try disable Tdr Watch preventing your PC crash and wait for operations ending.
Just set TdrLevel to zero and try to wait for 5-10 mins.
If it does not help you, maybe your problem relates on hard drive or ram memory, that causing pc freeze.
The settings that i’ve provided earlier works on RTX 3080 Ti. Maybe you hardware need more time to end all the operations or you have problems with your hardware =
I am sorry if it does not help you =\
TdrLevel
Specifies the initial level of recovery.
registryCopy
KeyPath : HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\GraphicsDrivers
KeyValue : TdrLevel
ValueType : REG_DWORD
ValueData : TdrLevelXxx (see the following table)
Where TdrLevelXxx can be one of the following values:
| Value | Meaning |
|---|---|
| TdrLevelOff (0) | Detection disabled |
| TdrLevelBugcheck (1) | Bug check on detected timeout; for example, no recovery. |
| TdrLevelRecoverVGA (2) | Recover to VGA (not implemented). |
| TdrLevelRecover (3) | Recover on timeout. This is the default value. |
My windows version
| Edition | Windows 11 Pro |
|---|---|
| Version | 22H2 |
| OS build | 22621.1413 |
I faced this error a few time, I ignored it and now this kind of stuttering is happening
https://youtu.be/bk1tFdacxHY
increasing TDR did not help my case
it is pretty much irrelevant given the amount of things i tried. all my previous experiments were with TDR time increased and check disabled
i tested metahumans on M2 mac, they work fine, but the hair/fuzz/lashes shaders dont work - making them invisible. no crashing.
RTX 3090 and AMD 5950X CPU here.
I finally seem to have fixed my random “d3d device removed” crashes by resetting my NVidia Control Panel 3D settings.
I had already increased TDR values because Substance software requires them. Still had crashes.
I gave up two months ago and was using DX11 in UE5. But, while trying to play Hogwarts Legacy this Sunday, I had way too many crashes and decided to hunt it once more. I was already on latest drivers as of April 23rd (531.68).
I like to fiddle with those NVidia 3D settings, so I thought maybe that’s how I messed up, because I hadn’t had these crashes for as long as many others. So it had to be something I changed recently, besides drivers, because I haven’t changed hardware since 2021.
I opened the NVidia Control Panel, restored defaults, and changed these settings:
After that, no more crashes in the game and Unreal has been stable in DX12. Maybe I just drastically reduced the amount of random crashes but, so far, none.
I really hope this helps someone else. I was getting sad by not being able to use the latest graphical features in the editor. These are really frustrating crashes, especially since Epic says it’s a driver bug, but only Unreal seems to be suffering from it. Weird, huh?
2 Likes
Dude, you have no idea. Now you know how I feel these days with this issue. Oh and I’ll definitely give it a try to see how this goes
1 Like
Just wanted to chime in and say i have exactly the same problems as you guys. Ryzen 3700X, 32GB of RAM, RTX 3080 Ti.
I tried all reasonable fixes under the sun. Increasing TDR delay does nothing, changing around drivers does nothing (i’ve gone back until the very first 5xx driver), changing/restoring NVIDIA Settings does nothing, disabling/uninstalling MSI Afterburner does nothing, messing with Game Mode and GPU Hardware Acceleration in Windows 11 does nothing, reverting back to Windows 10 does nothing. There is just nothing we can do to remedy this issues.
This is solely on Epic’s side of things as there are also many many crash reports out there using AMD GPU’s so a Driver/NVIDIA related bug can almost certainly be ruled out.
It has something to do with their DX12 implementation as Vulkan does not show any crashing issues (while still supporting Nanite and Lumen) even in Scene’s where VRAM and GPU usage is at max. I’ve created a test scene with different MetaHumans. DX12 - Crashes while loading the map, Vulkan 5 FPS but no crashes even after 2 Hours.
It seems to have something to do with their Garbage Collection while running in DX12 as the issues (at least for me) only happens when the editor free’s up previously used resources. I can hear the fans ramp up, memory usage drop for a bit and then comes the crash.
Over the last couple of months i’ve sent them around 4GBs of logs and crash reports but i have yet to see any response from them.
In it’s current state the engine is unuseable as almost anything can crash the editor at any time. Even if you just scroll long enough through the content browser IT WILL crash as soon as it triggers it’s own garbage collection.
Shader compilation also seems broken to some degree, it can’t be right that an empty (Blank) project creates around 5000 files in my DerivedDataCache folders, this simply kills ssd’s. My biggest project creates over 93000 files.
Edit: I should maybe also mention that my system is rock stable in over 300 installed games, passed a 12h prime95 stress test, passed 3h of furmark burn in testing without any errors and scores way above similar/comparable systems in 3DMark. So i can say with 99% certainty that my system is not the problem.
3 Likes
Nvidia just released version 531.61. I’m gonna give a try and see if solve the issue.
Otherwise, I’ll have to roll back to 528.49.
BTW, I saw many of you adviced to roll back to 517.40 but 528.49 was running fine on my configuration.
Edit : ==> still crashing with 531.61
Edit : ==> 528.49 ok so far
It helped a bit, there are less crashed.
So is it about power supply?
Having the same issue with MSI RTX 4070.
Is there a fix? anyone?
1 Like
its pretty apparent at this point that the fix should come from Epic, and no shenanigans and placebo will solve it by itself
1 Like
if you look better i am not only increaseing tdr delay, i am also changing
“TdrDdiDelay”=dword:0000003c
“TdrLimitCount”=dword:0000000a
tdrDelay will not help you as your gpu may crash by tdrLimitCount or TdrDdiDelay
You can disable tdrwatch by changing tdrLevel to 0. But your pc may freeze forever in that case if it cant finish operations.
anyway, i just wrote about my case how i fixed all my crashes and it really depends on hardware.
3 Likes
retried tdr and tdrddi delays with 60 and 600 values, same old d3d crash bug
1 Like
you forgot TdrLimitCount without this GPU still crash but no by delay. It will not have engough retries
Please read more carefully my messages.
There are 3 values you may change or change “TdrLevel”=dword:00000000
As we found , it helped only me 
But we need to try, maybe someone will fix his issues too
Windows Registry Editor Version 5.00
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers]
"TdrDelay"=dword:0000003c
"TdrDdiDelay"=dword:0000003c
"TdrLimitCount"=dword:0000000a
Maybe my windows version must be here too
| Edition | Windows 11 Pro |
|---|---|
| Version | 22H2 |
| Installed on | 25.03.2023 |
| OS build | 22621.1413 |
| Experience | Windows Feature Experience Pack 1000.22639.1000.0 |
greetings. thanks for the suggestions, unfortunately this does not help as well. it is apparent that these settings give the machine a bit more tries to recover, yet the root of the problem is not addressed, and the result is the familiar d3d crash window
siwtching to dx11 and performing same tests does not crash, so it is definitely the problem with dx12
Chiming in on this 3900x, 32gb, 2080ti. Tried most of the fixes above, nothing is working. I get DX12 crashes within minutes of loading up. Vulkan/DX11 no issues.
Hello, I’ve been having this issue for several weeks (see my earlier post above). Today one of my graphics cards (GTX 2080Ti) completely gave out. I’m wondering whether the card could have given out due to my changing Tdr settings in the registry, or anything else related to this issue? I’ve deleted the registry keys I created for now, and I still have one working 2080Ti, but is it possible that this ran my card into the ground?
i keep getting the D3D crash, 2 times in the last 3 days, untill 3 days ago i had never had a crash like this (1st 3 days of messing with UE5 were good, then bam, ICS crash cuz of some D3D bs …) same window popup as this person’s post too, more or less [different stuff shown in the file list but same window, colors & all) …
is there ANY fix to this AL ALL that doesn’t require fairly decent knowledge of how to work on-fix Software errors, etc &/or work on/replace Hardware? {or worse still, needing to either replace my Graphics Card with an RTX 4K Series or outright buy a whole new computer? …}
{running an RTX 3060Ti with 16Gb Ram, 8 Cores, … idr the rest right now X_X }