It’s a driver issue, same issues with a 2080 Super here.
This has been plaguing me and my RTX2080ti since I bought it in 2018. I don’t beleive it’s an Unreal specific bug as the same thing happens when I run RTX enabled games that use different engines. My VRAM is almost never maxed out when it crashes too. It’ll happile crash at 5Gb out of 11Gb that my card has. I guess there’s a chance these 1st gen RTX cards came with faulty VRAM. I’m at a complete loss as to how to fix it.
It happens with all hardware raytracing options disabled but only after importing a metahuman on my current project. It seemed stable even after importing a rather large 3D scan. But as soon as I bring a metahuman in and enable all the required plugin, it’s all over and the project crashes as soon as it opens while trying to compile the new shaders.
1 Like
Hi All! I red all the comments carfully and tired all the proposals, non of them worked.
Im just trying to start a new project with raytracing enabled, on DX12 with no luck.
Win 11. 32Gb RTX3060 12GB Studio Driver.
(I did not tired Vulkan neither Win10 cause i dont want to use them…)
Im running the latest Twinmotion on this machine with quiet large projects with minor problems, but im not shure what version of the engine it uses…
UPDATE!
It seems like the 517.40 Nvidia Studio Driver solved my issue!
It worth mention that I have the compiled “hung fix” .dll in place, what was mentioned above.
UE 5.0.3. Win11 RTX 3060
Cheers!
1 Like
I’m seeing this happen now in the Unreal 5.1 branch on AMD Video cards. This is clearly NOT an NVidia bug. I’ve seen a few crashes related to this in 5.0.3. and it really looks like a race condition.
Sometimes immediately after Unreal’s Garbage Collection runs UE crashes because the Texture got removed before the material. When this bug happens with the VS debugger attached, you cans see Unreal is trying to access invalid memory.
The whole GPU Crash or d3d device removed is just a symptom of this race condition as you can see in the video below in a blank project in UE 5.1… I’ve delt with a lot of race conditions in my career and when it smells like a race condition and looks like a race condition, it always has been. 
1 Like
Will just remember, the 5.1 branch from GitHub is not very compatible to use as much. If I was you, I would try to wait for the official release of 5.1 to come to the launcher soon. And see what any of that type of glitch bug is fixed for the hotfix. Whenever day they announce it
Same here, updated to the latest Nvidia Studio drivers, and now FINLLY able to enjoy DirectX 12 features in UE5 without the terrible crash. BTW, did not implement “hung fix”. Thanks for the update.
UE 5.03 Win11 3060ti
2 Likes
This is the correct solution
So far this is working for me. Will update with rage if it doesn’t.
thanks fixed for me too! from china
I have completely resolved my issue by using GPU underviolting without loosing performance (the main resolver) + Nvidia Studio drivers:
Here is the video:
1 Like
Had the following problem:
- with 5.0.3, 5.0.3 src and 5.1,
- RTX 3090
- Latest Win10 updates
- Latest NVIDEA drivers
Opening any UE5 project (new or loaded), as soon as I go into the menu bar, hovering over any icon there, the menu bar begins to flicker for around 5 seconds, then UE5 crashes. Crash reason given: EXCEPTION_ACCESS_VIOLATION
Temporarly solved it by adding “-d3d11” as a startup parameter, but this will give troubles with nanite and lumen…
Finally found out that it seems to be a strange settings problem.
Converting any UE5 project back to UE4.27 gives exactly the same behavior, even as I never had before any issue like this in 4.27.
So what helps?
- Start UE5 with “-d3d11” (so it don’t crashes)
- Open the UE5 project
- Go to settings, search for “Anti Aliasing”
- Switch the “Anti-Aliasing Method” to “None”
- Close the settings panel
- Open it again, search again for “Anti Aliasing”, and switch the “Anti-Aliasing Method” back to “Temporal Super-Resolution (TSR)” (or what it was before).
The project is now fixed, and I can open it without the “-d3d11” startup flag, with full lumen and nanite support.
Cheers
2 Likes
Have you found a fix for that? AMD card user here too.
DX12 haven’t been working for me at least since 5.x branch. Vulkan and DX11 are fine. The moment I try to move somewhat the viewport or load some objects on it, after few seconds, it crashes with Device removed.
Tried with the TdrDelay fixes - no luck.
I tried practically all of the suggestions here, including the solution found by @ClaudeSteiner, but none of them worked for me.
However, thanks to an unrelated Metahuman incompatibility issue between MetaHumans created with 4.x and 5.x, and an error in the dialog describing the issue, I had to recreate my entire project from scratch and migrate all of the content across to it.
I then went through all the project settings and ensured they were the same between the two projects. I still experienced the crash, but then I followed @ClaudeSteiner fix and this time it worked.
I don’t experience the device removed crash when opening levels in the new project. In the old project, when opening those same levels, I do experience the gpu device removed crash.
Recreating an entire project from scratch might not be practical and can be a lot of work, but if you did want to try this solution, you might create a new project and just migrate a single level across to see if the crash persists in the new project.
Thank you for the reply! Regretfully, toggling the Anti-Alias method didn’t do anything for me. Even on new example projects, UE crashes on dx12.
FYI to others. I was having this issue (3080Ti Win11, UE5.1). Tried a number of things, including DDU’ing my drivers, changing recommended windows timeouts, etc. In the end it turned out somehow a setting in my BIOS got changed that was messing with my RAM memory voltage/timings.
I discovered this by running the built in “windows memory diagnostic” which failed and claimed I had hardware errors. This lead me to discovering and setting the BIOS settings back to reasonable values. I re-ran the memory checker and it successfully completed multiple passes. Haven’t had the issue since.
Hi everyone
What fixed this for me was going into (nvidea control pannel/Manage 3d settings and clicking on Restore and then apply). Not sure if this works for anyone else, but i did for me.
Hello, I’ve been getting this crash report occasionally (maybe 3 or 4 times a day) and I’ve ended up reading through this thread a few times, here’s what I’m using:
|-| The setup I’m using is Unreal 5.1 (NOT built from source)
|-| Windows 10 Pro, 2080Ti, plenty of RAM and disk space
|-| Latest Nvidia Studio drivers installed
|-| I’m not using any overclocking or GPU monitoring software, it’s not something I have used before.
|-| I’ve set the registry keys as recommended, TdrLevel = 3, TdrDdiDelay = 60, TdrDebugMode = 3, TdrDelay = 60
|-| I’ve disabled GSync in the Nvidia control panel
I’m going to try using dx11 for now, but what would you try next to get dx12 working?
So…how do you fix this then?
Ok, let me add my side of things too…
- It doesn´t only happen on NVIDA cards, there is also a long thread on AMD cards with the same issue.
- I had this issue on my old 2080 and now on my new 3080, so its not related to these cards either.
- my colleague, also 3090, doesn´t have this issue.
- I´m on windows 10 (21H2), so its not a windows 11 Issue
- Registry hack didn´t work.
- installing either studio or gamedriver didn´t work.
- Upgrading driver to newest version didn´t help (currently 528.49)
- Couldn´t use the “hung dll” fix, as I got an error message, that my system doesn´t meet the requirements for Unreal 5.1. Nanite seemed to be broken, so I´m assuming it just made Unreal revert to Direct X11, which seems to work for everyone here?
Can´t use that though for any of my projects, as all rely on nanite… - Limiting the frame rate to 30fps seems to fix it. Since unreal 5.1.1. I´m having the issue though, that whenever I open ANY editor windows, fps drop to 10-15fps.
I experimented a little, and I can probably go up to 50fps, which then means a drop to 20-25fps when opening windows.
Anything higher (80fps was taking the longest to crash…) makes it crash, so I´m assuming the issue is related to this. - The aliasing fix did nothing for me.
- Disabling G-Sync did nothing for me.
- I checked with GPU-Z and my card isn´t overclocked. Neither was the 2080 I used before.
- Reinstalling drivers with DDU also didn´t help
- All the other issues regarding various metahuman or other things don´t apply, as I don´t use any of these and still get crashes.
- Disabling hardware raytracing also didn´t fix it.
The only things left to try for me are:
16. Checking for some Bios Setting being abnorm.
17. Installing MSI Afterburner and doing the low voltage thing.
This is clearly nothing EVERYONE experiences, but its clearly a very common thing, so we should be able to figure this out, if EPIC games isn´t able to fix it.
For now I´ll see how long I can stay crash free with 50fps and report back if the last two fixes did anything…
2 Likes