UE5 Preview 1 crashes gpu. DXGI_ERROR_DEVICE_HUNG

This is what the log looks like… thanks if anyone takes the time to have a look.

I’ve got a 2080 SUPER with newest drivers.

Thanks!

[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Status: PageFault
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Failed to get Aftermath stack data
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Faulting address: 0x000000370c9f5000
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Faulting resource dims: 0 x 0 x 0
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Faulting result size: 0 bytes
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Faulting resource mips: 0
[2022.03.05-02.38.45:806][ 73]LogRHI: Error: [Aftermath] Faulting resource format: UNKNOWN (0x0)
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: PageFault at VA GPUAddress “0x370C9F5000”
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: Last completed frame ID: -1 (cached: 1069) - Current frame ID: 1072
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: Logging all resource enabled: No
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: Found 0 active tracked resources in 16.00 MB range of page fault address
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: Found 0 active heaps containing page fault address
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: PageFault: Found 0 released resources containing the page fault address during last 100 frames
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: [GPUBreadCrumb] Last tracked GPU operations:
[2022.03.05-02.38.45:806][ 73]LogD3D12RHI: Error: [GPUBreadCrumb] 3D Queue 0 - Begin: Frame 1069 - Scene - Lights - DirectLighting - UnbatchedLights - MHC_LightingPreset_Split_JD.KeySpotLight - ShadowBatch - MHC_LightingPreset_Split_JD.RectLight - RayTracedShadow (spp=3) 695x770
[2022.03.05-02.38.45:809][ 73]LogD3D12RHI: Error: [GPUBreadCrumb] Compute Queue 0 - No Data
[2022.03.05-02.38.45:820][ 73]LogD3D12RHI: Error: [GPUBreadCrumb] Copy Queue 0 - No Data
[2022.03.05-02.38.45:831][ 73]LogD3D12RHI: Error: DRED: No breadcrumb head found.
[2022.03.05-02.38.45:832][ 73]LogD3D12RHI: Error: DRED: No PageFault data.
[2022.03.05-02.38.45:832][ 73]LogD3D12RHI: Error: Memory Info from frame ID 1070:
[2022.03.05-02.38.45:832][ 73]LogD3D12RHI: Error: Budget: 6808.50 MB
[2022.03.05-02.38.45:832][ 73]LogD3D12RHI: Error: Used: 3539.84 MB
[2022.03.05-02.38.50:309][ 73]LogD3D12RHI: Error: pResource->Map(Subresource, pReadRange, reinterpret_cast<void**>(&pData)) failed
at D:\build++UE5\Sync\Engine\Source\Runtime\D3D12RHI\Private\D3D12RHIPrivate.h:1333
with error DXGI_ERROR_DEVICE_REMOVED with Reason: DXGI_ERROR_DEVICE_HUNG

[2022.03.05-02.38.50:317][ 73]LogWindows: Error: Error reentered: Fatal error: [File:D:\build++UE5\Sync\Engine\Source\Runtime\D3D12RHI\Private\D3D12Util.cpp] [Line: 859]
hr failed
at D:\build++UE5\Sync\Engine\Source\Runtime\D3D12RHI\Private\D3D12Resources.cpp:629
with error DXGI_ERROR_DEVICE_REMOVED with Reason: DXGI_ERROR_DEVICE_HUNG

1 Like

I have the same issue!

Regardless if i use DX11, DX12 or Vulcan, switch AA Types or and activate Virtual Textures/ Shadows,
it still chrashes randomly in complex and relativly simple scenes(Sky box and lowpoly meshes, Sun and Skylight), cant work for more than ca. 10min a time.
It affects all Projects running in UE5 preview 1.
Does anyone has an other solution or do we have to wait for Epic to fix this.

GTX 1080TI Strix - 11GB - also newest drivers
i7 - 5820K - oc: 4GHz
M.2 SSD and X99ASUS Motherboard
My Hardware is sufficent, therefore im 100% sure its a Bug, in UE5 EA there was no such Problem!

Best greetings,
-Eric

Yep, I get about 5-10 minutes before it crashes and I too had no issues with Early Access either.

1 Like

I cleanly reinstalled UE5 prview 1, now its even worse after 10-20 seconds it chrashes in the default open world map and/or empty map. Also for the first time UE5 EA has crashed on me, as if UE5 prview 1, has affected the other version. Both crashed in blank projects.

I subbmited a BUG Report.
100% sure it is an Engine Problem!

The editor still crashes in Preview 2 but now it only takes about a minute before it crashes.

I’ve tried various drivers… no luck.

Preview 2 is essentially impossible to use.

Same here. No such problems in UE5 EA.

The weirdest part is no one else seems to be complaining about this other than the few that replied to this thread.

Hmm.

I switched to Vulcan and it seems so far to crash less. Also it crashes more if I use one of the Open World levels vs the standard level, the one that just says Level is still world partition, so use the one that says Standard. I think it might be related to World Partition. Let us know the results, maybe we can help narrow it down.

Specs:
Windows 11
Rtx 2070 Super

I turned off Virtual Textures and I’m getting far less crashes.

Just confirming that VT + Metahumans seemed to be the issue.

Crashes are rare now that I turned VT off in Preview 2.

Same problem here. I just wanted to try the Lyra example project and then after a few minutes:

[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Status: PageFault
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Failed to get Aftermath stack data
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Faulting address: 0x00000003e840c000
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Faulting resource dims: 1160 x 1024 x 1
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Faulting result size: 4784128 bytes
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Faulting resource mips: 1
[2022.04.06-20.03.04:867][476]LogRHI: Error: [Aftermath] Faulting resource format: R11G11B10_FLOAT (0x1a)
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: PageFault at VA GPUAddress “0x3E840C000”
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: Last completed frame ID: -1 (cached: 24494) - Current frame ID: 24497
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: Logging all resource enabled: No
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: Found 0 active tracked resources in 16.00 MB range of page fault address
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: Found 1 active heaps containing page fault address
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: GPU Address: “0x3E6000000” - Size: 128.00 MB - Name: TransientResourceAllocator Backing Heap
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: PageFault: Found 0 released resources containing the page fault address during last 100 frames
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: [GPUBreadCrumb] Last tracked GPU operations:
[2022.04.06-20.03.04:867][476]LogD3D12RHI: Error: [GPUBreadCrumb] 3D Queue 0 - Begin: Frame 24473 - Scene - DiffuseIndirectAndAO - LumenScreenProbeGather - TemporalReprojection 1159x859
[2022.04.06-20.03.04:874][477]LogD3D12RHI: Error: [GPUBreadCrumb] Compute Queue 0 - No Data
[2022.04.06-20.03.04:885][477]LogD3D12RHI: Error: [GPUBreadCrumb] Copy Queue 0 - No Data
[2022.04.06-20.03.04:896][477]LogD3D12RHI: Error: DRED: No breadcrumb head found.
[2022.04.06-20.03.04:896][477]LogD3D12RHI: Error: DRED: No PageFault data.
[2022.04.06-20.03.04:896][477]LogD3D12RHI: Error: Memory Info from frame ID 24495:
[2022.04.06-20.03.04:896][477]LogD3D12RHI: Error: Budget: 7275.00 MB
[2022.04.06-20.03.04:896][477]LogD3D12RHI: Error: Used: 4483.59 MB
[2022.04.06-20.03.20:234][477]LogD3D12RHI: Error: pResource->Map(Subresource, pReadRange, reinterpret_cast<void**>(&pData)) failed
at D:\build++UE5\Sync\Engine\Source\Runtime\D3D12RHI\Private\D3D12RHIPrivate.h:1387
with error DXGI_ERROR_DEVICE_REMOVED with Reason: DXGI_ERROR_DEVICE_HUNG

UE 5.0
Windows 11
Nvidia RTX 3070

Hello! This issue I’ve found to be due to overclocked GPUs. In my case, the RAM was overclocked by default on my Zotec 1080ti amp extream edition. Running nvidia in a Debug mode helped (it sets your clocks to stock.) But my GPU Ram needed a bigger underclock, I currently have it at the lowest the FireStore Zotec software will allow at -100Mhz. I underclocked the GPU core and still had the hangs, and the core should draw more POWER than the RAM. So it should not be a PSU issue, but that could also be there… as we are basically saying the GPU fails and reboots (hangs on a frame) when running at its limits, which overlocked cards are closer too (their breaking points/lints and reboots.) PSU failures tend to be more regular, not freak. undersized could be this, but again, my ram was the issue. GOOD LUCK - Have Fun!
Micah #roadlessthegame

Thanks! I underclocked my 2080 SUPER by 100 mhz and so far it’s MUCH more solid.

I didn’t touch the memory speed but I’ll try that if it acts up again.

Cheers!

1 Like

Have same problem.
Use Enable Hardware Ray Tracing in Project Settings fixed a bug for me

Getting same intermittant crash on UE5, DX 12, rtx 3080-ti, latest drivers

Same project doesnt crash at all on various other machines.

Will try DX11/Vulkan but hoping for a better solution in future / enabling Hardware RT

So - as strange as it sounds… removing all the groom assets off the pawn ( I am using one of the stock metahuman characters) seems to have fixed the crash for me. For some reason, Groom seems to be causing it. No idea why groom would cause it only on the machine with 3080ti. Other machines with 3090, 3070, even 3050 ti tablet are all ok with it.

Try to execute the console command t.MaxFPS 60, it helped me.