UE5 gpu crashed or d3d device removed

i just wanna get to use the stuff i found from Unreal Sensei & Smart Poly (including his 60$ Game Development course Lessons i bought) to learn a buncha stuff & finally get to a point i can maybe, just MAYBE build AN incarnation of atleast 1 of my game concepts
… But Unreal keeps crashing, corrupting Project Files (basic as they are) & thus preventing me from actually getting anything useful outa the stuff :sob: :face_with_head_bandage: :skull_and_crossbones:

This issue has somewhat faded away for us… No idea why, but perhaps this might help someone as I know how frustrating it can be. Try removing all UDIM textures.

We’re now using 5.2, and the only major changes we’ve tried since updating is that we removed all UDIM texture sets (we only had a few). This seems to have removed our D3D crashes, though messing with some more intensive processes in-engine can result in the regular crashes and Unreal crash reporter popup.

idk what UDIM Textures even are specifically sooo … (& if the crashes are STILL happening anyway when digging into the heavier systems/Tools, then the problem will still persist & make it impossible for me to ever accomplish ANYTHING unless i somehow manage to afford a PC with a 4K Series RTX Card … :face_with_head_bandage:)

if you dont know it, quite possibly you dont have it (you need specifically enable VTs first, and then have assets with UDIMs)

also, people with “big” GPUs reported that problem as well, so you might get same problems even if you get a new card :face_exhaling:

UPD:
there are new NVidia gameready drivers, claiming to have bugfixes as stated here:

Fixed Bugs

  • [RTX 30 series] Application crash may be observed while playing Battlefield 2042/The Last of Us/Unreal Editor [4013983]

Gonna throw at it my typical crashing scenarios and see if it got any better. Anyone else feel free to get new drivers and chime in :melting_face:

UPD2:

Seemingly it does indeed improve the d3d crash. Throwing my usual d3d crash scenarios at the updated driver is stable enough. I have not yet tested it with big UDIMs, but enabling RT + Shadows + all the whistles is still stable enough, so that is definitely an improvement. The 531.79 driver seems to be the way to go.

Please share your own experiences, I will be (lazily) testing it more over the week :face_in_clouds:

I am still getting these crashes, tried game drivers, studio drivers, 532.04 developer drivers, tdr, every bios setting under the sun. 4090, 5950x, 128gb ram.

Tbh, recent win 11 updates seem to have made it worse, I’ve tried with hardware accelerated gpu scheduling turned on and off, gsync on and off.

Deleting cache from appdata folder seems to have some beneficial effect for a while, sometimes.

Tempted to splurge on 7000 series and x670 just to see if it’ll help.

2 Likes

I have a RX 6650 XT and same is happening for me. Downgrade to windows 10? Seriously is this the solution. Have to reinstall all my stuff again.

Nevermind, turned out to be my PCIe riser cable that had given up the ghost, no crashes since plugging the GPU directly into the motherboard. Funny how it wasn’t affecting any other games or apps though.

I’m regret to say that 4K Series RTX card won’t fix this crash. Mine is 4090 with 24G vram, and I got even more crashes than my old 3070ti. Now this is Ureal just being unreasonable…

1 Like

Any more results from testing that driver version?
I haven´t had a crash in quite a while and I can´t quite remember what level gave the constant crashes, so I could at least check if it was just that particular level (although I´m pretty sure its not a performance issue, because I´ve done a lot more taxing scenes since without issues).
Last working solution I remember was limiting max frame rate but I just checked and apparently I turned it off after last round of testing.
We also upgraded to Unreal 5.2, can´t quite remember if thats when the crashes stopped, but I would have probably noticed that…
I´m currently on 528.49.

UE5 5.2, RTX 3090 + Ryzen 9 5950X
536.23 Game Ready drivers, with Studio happens the same.
I’m undervolting with MSI Afterburner an did the Tdr delay regedit setting
I keep having continuous crashes… “DXGI_ERROR_DEVICE_REMOVED with Reason: DXGI_ERROR_DEVICE_HUNG”

Looks like it’s just random, also with LyraStarterGame, a part from my current project.

Help!!!

Edit:
It looks like it was the PCI-E for my vertical mount GPU. After setting type 3.0 instead of 4.0 or Auto in my BIOS, it has been completely stable.

I have the same GPU/CPU and I rarely get the hung error.
When I first upgraded to Win11 it happen constantly in 5.0.
I tried all the tips found in the forum (regedits, studio drivers, etc.)
I even edited the engine’s DirectX12 cpp files and it stopped crashing, but it didn’t work for other people.
I have the PC under my desk and the heat sucked, so a while back I set the Power Limit to 80% in afterburner and set to start with windows to lower the temp and use less power when playing games (no need to use 400watts playing BF2042) :crazy_face:
In 5.2 I never edited the DirectX12 cpp files like I did in 5.0/5.1. So I wonder if the power limit is what was really preventing the crash in Unreal.

Ok, so, I am back, got a new scene now that keeps crashing on me.
This time its metahuman+transparent material in the same scene.
CBA to figure out which part of the scene crashes it.
Tried your fix again, this time with all three values as suggested, but as others have pointed out: This only delays the crash, it doesn´t fix it.
If you don´t crash anymore since you added this, its either related to your hardware, or you ALSO changed something else at around the same time and THAT fixed it…
OR you simply haven´t worked on a scene that causes that crash (which happened to me for several weeks/months now apparently).

Ok, updated to the latest version of gameready drivers, 536.23.
Crashes at around the same frame if I scrub the sequence. It only crashes,
when I have an animated camera in the sequence or if I move around the scene with the viewport cam.

But while diddling around with the scene, I thought I had found another “workaround” :
I have my engine scalability settings set to “Epic” by default, but for some scenes I also need to set at least “Textures” and “Shaders” to “Cinematic”.
(For example Metahuman textures on epic look like complete garbage for me).

So, without changing anything from before, I set the scalability to “high” and it didn´t crash.
Next I switched everything to epic and it didn´t crash.
So then I first set Textures and then Shaders back to cinematic (same settings as before), but it didn´t crash anymore.

Unfortunatelly, while this worked twice in a row, third time it crashed again.
I didn´t keep testing much further than this, everything high and only textures on cinematic still worked, everything on epic crashed again.
So, if you have a scene you can crash reliably with your current settings, you could
do some further testing, but I´m not sure thats going to help everyone, scalability settings might still be scene/hardware dependent.
Scene sits around 60fps normally with 17ms and about 6gb Vram on a 3090.

The ONLY reliable “Fix” from all 155 posts so far is to limit the frame rate in the driver.
For me that is to about 58 fps, which makes taxing scenes run around half of the fps in editor
Which is still fine to work with in most cases, but in some scenes, the fps drops down to 20 or even 15 fps, which it wouldn´t do without the frame limit.
Ms also goes up from around 17ms to around 50ms.

Dear Epic Devs…can you please look into this?
I love Unreal so much, but this has been going on for way to long and for way too many people.
If you cannot reproduce the error, I´m happy to provide a test project.
And I know just because dozens of people experience the same issue, it still might be related to random hardware or hardware-constellation problems.
But maybe you could at least chime in with something to try to pin point the issue some more…?

sitting at 535.98 (latest Studio), all works reasonably stable

Hi, thanks for you responce!
After sometime i also noticed that unreal engine also RAM and GPU Memory sensetive and may call “d3d device removed” when you dont have enough RAM or VRAM(mostly VRAM)

I got d3d device removed when tried to render scene that needs 45 GB~ Video Memory

i dont have much, before. i only got 32 GB Ram and 12 GB VRAM that.
So i only can get 28 GB VRAM (MaxRAM/2 + VRAM) . Also you cant just swap Shard Video Memory, it need only Physical RAM

After upgrading RAM to 128 GB, now i dont have any d3d device removed while working with really big scenes that can consume really big amounts of VRAM or RAM (excluding memory leaks, that can still crash your engine, but i did not got exactly d3d device removed)

How to determine that the problem is in RAM?
Try to monitor you Ram in TaskManager and if you get crash wher your current RAM = MaxRam then its yours’

How to determine that the problem is in RAM?
The same as RAM. I got crash only when VRAM and SharedMemoery was full.

If you crash not in this cases so the problem can be deeper in you hardware setup.

Sorry, didn´t explicitely mention this, but my Vram was sitting comfortably at 6GB when it crashed, on an rtx 3090, which has 24GB, so I really don´t think Vram is the issue.
It might spike for some unknown reason, something like cache being emptied before loadinga scene etc, but thats not something I would know, only devs could answer that.

I have never heard that normal Ram would cause this, but I have 64GB and I´m pretty sure it also wasn´t maxed out when it crashed.

I might wanna dig deeper into the new profiling tool though, maybe it helps figure this out…

Try setting the motherboard’s PCI-E 3.0 instead of 4.0 or Auto in the BIOS, it worked for me, although that was because I had a GPU vertical mount.

I have AMD Radeon 6800 XT and Windows 10. Tried updating drivers, but that didn’t help.

I found 2 solutions:

  1. switching to DX11 renderer
  2. disabling the HW RayTracing in Project Settings.

I went for the 2nd solution.

Had non stop D3D crashes with a project using 5.3.0P1 and a 4090 with latest drivers and extra regedit settings.
Using the command option ‘-gpucrashdebugging’ found a reference to ‘particles’ at the end of the log file. After removing the 3dGasIntro campfire (NiagaraFluid plugin) from the project, the non-stop crashes are gone. Crossing fingers…

See How to Fix a GPU Driver Crash When Using Unreal Engine | Unreal Engine 5.0 Documentation

Edit: in another project I also got many crashes. It seems to get solved by opening the Nvidia menu and reset it all to default settings.