DXGI hang only in Battlefield 6.

I seemed to have solved the issue for me at least. I have disabled my integrated graphics (iGPU) in the BIOS and haven't crashed in the past few days, can now run an overclock on my gpu. Hopefully this works for others

Currently I have quite severe underclock+volt on my 9070xt -150mhz and -60mv with stable 60% fan speed to alleviate the crashing, still happens, but atleast I can play 2-3 rounds of breakthrough. Also have -dx11 in steam launch options and GstRender.Dx12Enabled 0 in PROFSAVE_profile, not sure if that has any impact or not. Two main pipelines causing the crashes are ShaderProgram_TerrainQuadIndexTessellation and ShaderProgram_GpuMeshScatteringCull_IndirectDraw in LodFadeStencilMask render pass.

Update. I can load into the training room but the game will crash within 2-10 seconds. If I reduce mesh quality to medium I can hang out in the loadout menu.

Sloth_Froth​ Hey, moved your thread to this one as we are keeping more of these DXGI errors together for visibility and steps that may help you out and get that resolved on your end.

Thanks

Battlefield 6 crashes fails with a DirectX error. Game loads to root menu, but when I try to load into a game or navigate to the loadout menu the game crashes.

System:

GPU: NVIDIA GeForce RTX 5090

Motherboard: Gigabyte AORUS X570 Elite, AM4

BIOS mode: UEFI

Secure Boot: Enabled

Memory Integrity/Core Isolation: Disabled

Errors seen:

1. DirectX function "GetDeviceRemovedReason" failed with DXGI_ERROR_DEVICE_HUNG ("The GPU will not respond to more commands"). GPU: "NVIDIA GeForce RTX 5090", Driver: 61047.

2. DirectX function "Hang Detected in waitForFrameFences. Timed out after 8000 ms. Waiting for SignalId: 24288. Last Signaled Fence Value: 24289. Unsigned Fences: ("Graphics" stopped at 24287)," failed with DXGI_ERROR_DEVICE_HUNG ("The GPU will not respond to more commands"). GPU: "NVIDIA GeForce RTX 5090", Driver: 61047.

Path shown: D:\dev\TnT\Code\Packages\RenderCore2\RenderCore2\dev\source\platform\Dx12\SubmissionManagerDx12.cpp

Also observed in Windows Event Viewer:

Source: nvlddmkm

Event ID: 153

Troubleshooting already performed:

- Full DDU driver cleanup and reinstall

- Tested multiple NVIDIA drivers, including 610.47 and 596.49

- Rolled back drivers

- Reinstalled Battlefield 6 on a different ssd

- Verified/repaired game files

- Cleared NVIDIA shader caches

- Cleared Battlefield 6 settings/cache

- Installed DirectX runtime

- Repaired EA AntiCheat

- Confirmed Secure Boot is enabled

- Confirmed Memory Integrity/Core Isolation is disabled

- Disabled Hardware Accelerated GPU Scheduling

- Disabled overlays including Game Bar / capture features

- Tested low graphics settings

- Disabled DLSS

- Disabled Frame Generation

- Disabled Reflex

- Tried FPS cap, including 60 FPS

- Tried fullscreen/windowed-style display changes

- Disabled HDR

- Disabled G-Sync

- Tested with single monitor

- Set refresh rate to 60/120 Hz for testing

- Disabled XMP/memory profile in BIOS

- Forced PCIe slot configuration to Gen 4 in BIOS

- Disabled Re-Size BAR

- Tested with GPU underclocked

- reduced power limit

- negative core clock offset

- negative memory clock offset

- Installed/updated AMD chipset drivers

- Updated/checked motherboard BIOS settings

- Added TDR registry keys for testing:

- TdrDelay = 10

- TdrDdiDelay = 20

This changed the error to the more specific waitForFrameFences timeout, but did not fix the issue.

Other notes:

- Other games run normally (007 first light, Silent Hill 2 remake, Final Fantasy 7 remake)

-I own game on Steam

Not sure fresh user works like fresh windows install. If files are corrupt they are corrupt on both users, anyway I've tried both fresh windows install and new user. Also according to EA_Barry I tried 26.3.1 Adrenalin Drivers again with full stock settings on all hardware except EXPO enabled. 

Summary from chatbot:

• Fresh Windows user account: no effect, crash persists
• Driver 26.3.1 (EA CM recommended): same crash signature
• ShaderProgram_TerrainQuadIndexTessellation / LodFadeStencilMask reproduced for the third time across different sessions and driver versions
• Both Graphics and Compute queues stalling simultaneously confirmed again
• Hardware telemetry shows no thermal, power, or PCIe anomalies preceding crash

S4NTZU​ It comes down to corrupted Windows files, typically. A new (let's say Amdin) user being created is like a fresh Windows install. It updates, repairs, and replaces any missing files to a kind of factory default state. So, anything that is missing, like a DLL file, for example, is fixed. It's a way to not have to reinstall Windows completely and lose everything.

Just fresh user, not fresh windows install? I don't understand why that would fix it but sure, good for you.

I've tried that aswell. Went down as low as -200 on the core clock and still got the error.