I feel like this is entirely an UE5 issue that they haven’t solved yet. I never have any crashes with any other app or game. Only UE5 with DX12 and Lumen active makes my GPU crash
Thanks, haven´t heard of that one before. unreal 5 uses TSR by default though, not MSAA and FSXA is the only other option
By “they” I assume you mean other users, not epic staff themselves, right?
Not pointing fingers here, but this IS a pretty long thread and no matter how many of us just suffer from unforeseen hardware compatibility issues or something like that…
the devs have been awfully quiet here.
Would be nice to hear something from their side on this issue every once in a while…
I haven´t had a crash in a while though, so I haven´t been looking into that more, even disabled the FPS limiter in the nvidia control panel.
I really really wish I had saved a map that could consistently crash it…
I’d say that it’s a combination of high impact features such as DX12, lumen, virtual shadow maps, nanite, groom etc. Any of those in combination, for some period of time cause crashes. I’m on the latest nvidia drivers with a 4090 and I have never experienced any crashes with any other engine nor game nor DCC app, ever. UE5 on dx12 is the only app that makes my GPU crash, yesterday I had about 15 crashes in a period of 6 hours. So it all points to this being exclusively an UE5 issue related to the way that the engine hogs resources.
1 Like
Just a quick update, had some issues with transparency in a scene and temporarily tried checking “Enable order independent Transparency (Experimental)” again and boom, had the GPU crash when rendering with MRQ, first time in months.
Unchecked it, worked fine again.
So, just to reiterate:
- Fixing framerate in Nvidia drivers
- Unchecking the transparency thing in project setting.
Until there is a definitive answer from epic staff, these seem to be the most reliable “fixes”.
Reverting to Directx 11 is NOT a fix/workaround, IF you need nanite/lumen…
I’m having this same issue with GPU crashes, and the dump always seeming to indicate transparency. I have order independent transparency on and have not tried to turn it off.
However, I found myself in a load-bearing printf situation somehow. Specifically, launching Unreal Editor with -d3ddebug makes it not crash.
I’m also facing same issue whenever. I start metahuman. boom GPU crashed or D3D device removed. i have an AMD RX 6600. Any one have any solution.
This worked for me! Thank you so much
Thank you very much! This reply really helped me a lot! Best wish to you! 
Hm, thats interesting, I´m gonna try and see if I can reproduce teh crash once more with order independent transparency on AND -d3ddebug…
I’d like to note that since that I’ve updated to tweaked the hell out of my project settings regarding how lighting is handled, studio drivers, rebooted, and I’ve had 0 GPU-related crashes since then (without -d3ddebug). Thing is, order independent transparency is off now, so that is a problem.
Thing I noticed with -d3ddebug is that at times (especially when anything transparent was being rendered, so just looking in the general direction of transparent objects even when obscured by walls) my message log would be obliterated by spammy error messages, so I presume that d3ddebug prevents the crash altogether before it can happen by detecting the issue. So all I’ve really learned to confirm with this was that it is indeed something transparency related (though that part was already hinted at by the crash logs, but GPU stack dumps could’ve been unreliable). The avenue seems mostly unexplored as googling the messages brought up nothing even close to what I saw, so maybe getting the info to Epic could result in some fixes.
So unfortunately, it is most likely that order independent transparency is currently not stable enough to rely on.
1 Like
I’m on a RTX 3070 with 8 GB VRAM. I fixed the d3d issue in 2022 with the registry tdr values turned to 60. After a week of experimenting with hardware raytracing the problem has reoccured.
Resetting the 3070 did not work, nor did changing the registry values to 120 or doing a clean install and even trying out older nvidia studio drivers. The compiled DX12.dll trick didn’t work either. Diagnostics show RTX (run dxdiag) and RAM (cmd mdsched) all ok.
Now I’m pondering on trying DDU, a Windows 11 factory reset or downgrading to Windows 10.
As of the moment no projects at all (even old 4.27 stuff, very basic and projects using DX 11) will last longer than 60 seconds if they don’t crash before that… Guess I’ll have to wait for this thread to get marked as resolved before using Unreal ever again 
Update drivers to Nvidia Studio is fix for me!
Thanks!
Hi, recently I had the same problem and I searched everywhere around when I came to try the remote renderer instead of the local renderer and guess what, it’s basically not crashing anymore. I can render 4K images without a crash for some reason, whereas the local one was crashing even on 1080p. I hope it can help some of you!
1 Like
So I have been getting the following dreaded messages on one of our levels. Didn’t do it in my test maps, asset gyms or anywhere else. So it was becoming a real pain in the neck.
Assertion failed: bAllowUAVs || (!bAllowUAVs && Counts.NumUAVs == 0) [File:D:\build++UE5\Sync\Engine\Source\Runtime\D3D12RHI\Private\D3D12Util.cpp] [Line: 1188]
It seemed impossible to pinpoint what was causing the issues. Was it my 14900k bios settings? Bad memory? Anything hardware related?
I updated bios, set everything to default and cleaned up my settings. Even took out 64gb of ram to drop from 128 to 64 and go in xmp mode just to test. Still no fix.
Latest nvidia drivers, clean install.
Clean install of UE5.4.
Nothing worked.
Finally got another message after clean install
and that lead to one of my coworkers suggesting to change my DDC paths and force my system to re-cache everything.
Turns out (after a long process of going through everything) that seems to have worked.
I’m back up and running. Not sure if this is the fix for everyone else but it is def worth a shot. It looks like at some point when loading or doing something data got corrupted somehow and it wasn’t getting cleaned up. Or well that is my best guess as an artist.
Anyways. Hope this information helps.
1 Like
I had the same problem this week.
I have tried nearly everything., even reinstalled windows, updated Bios, regedit entries, older graphic drivers. Nothing helped.
Finally i loaded the default settings in the bios.
And now it works again.
Perhaps this is useful for somebody.
1 thing I found is that the rivia tuner show on screen display can cause a crash