Unreal Engine Exiting Due to D3D Device Being Lost: Fix

On this page

“Unreal Engine is exiting due to D3D device being lost” means the game lost its connection to your graphics card, and the error code after it tells you why: 0x887A0006 'HUNG' is a GPU timeout, 0x887A0005 'REMOVED' is a full loss of the card. It is a PC-side problem — driver, overclock, power or heat — not a game bug, which is why the same crash appears across dozens of Unreal Engine games. Read your code first, then work the fixes below in order.

Unreal Engine Exiting Due to D3D Device Being Lost: Fix

The full message usually looks like this:

LowLevelFatalError [File:Unknown] [Line: 200]
Unreal Engine is exiting due to D3D device being lost. (Error: 0x887A0006 - 'HUNG')

The line number varies by engine version. The part that matters is the code in brackets.

Step 1 — Read the error code

Unreal Engine doesn’t generate this failure itself. It reports the error Windows’ DirectX layer (DXGI) handed back, so the code maps directly to a DXGI error:

Code in the messageDXGI errorWhat happenedGo to
0x887A0006 - 'HUNG'DEVICE_HUNGGPU took too long; Windows’ TDR reset the driver (~2 s)DXGI_ERROR_DEVICE_HUNG fix
0x887A0005 - 'REMOVED'DEVICE_REMOVEDWindows treated the GPU as unpluggedDXGI_ERROR_DEVICE_REMOVED fix
0x887A0007 - 'RESET'DEVICE_RESETDevice reset, often by bad commands or a driver faultFixes 1–3 below
0x887A0020DRIVER_INTERNAL_ERRORThe driver itself failedFix 1 (clean driver)
0x0 - 'S_OK'none — a known UE 4.26/4.27 caseOverlay/MPO conflict, iGPU + NVIDIAFix 5

HUNG is by far the most common. If you see it, the dedicated DEVICE_HUNG guide goes deeper on the TDR timeout; this page covers what is specific to Unreal Engine games.

Fix 1 — Clean-install your GPU driver

A corrupted or half-updated driver is the most common cause and the cheapest to rule out.

  1. Download the current driver for your GPU from NVIDIA, AMD or Intel.
  2. Boot into Safe Mode and run Display Driver Uninstaller (DDU) to remove the old driver completely.
  3. Install the fresh driver and reboot.
  4. Clear the shader cache so the game rebuilds shaders against the new driver.

If the crash started right after a driver update, install the previous driver version instead — a regression in one release is a real possibility.

Fix 2 — Remove every overclock

Unreal Engine 5 games push GPUs hard, and an overclock that is “stable” in other games often isn’t here.

  • GPU core and memory: reset to stock in MSI Afterburner or your vendor tool. Factory-overclocked cards can also be unstable — try a small negative core offset (−50 to −100 MHz) as a test.
  • RAM: if your code is REMOVED, briefly disable XMP/EXPO to run at JEDEC speed. Unstable memory can cause the GPU to drop off the bus. See how to enable XMP or EXPO to re-enable it afterwards.
  • CPU: undo any undervolt or PBO/curve tweaks during testing.

If the crash stops at stock clocks, add your overclock back gradually to find the unstable setting.

Fix 3 — Rule out power and heat

The GPU losing contact mid-game is often a hardware-level brownout or thermal event:

  • Use separate PCIe power cables for each 8-pin connector instead of one daisy-chained cable, and reseat any 12VHPWR/12V-2x6 connector fully.
  • Check your PSU has headroom for your GPU’s peak draw — transient spikes are much higher than the rated TDP.
  • Watch GPU hotspot and memory temperatures in an overlay. Sustained temperatures near the card’s limit can trigger the crash. Clean dust, improve airflow, or cap your frame rate to lower the load.

Fix 4 — Switch DirectX version and cut overlays

These are Unreal-specific workarounds that help when a driver bug affects only one API or when an overlay is involved:

  • Try the other API. Many Unreal Engine games accept a launch option — in Steam, right-click the game → Properties → Launch Options — of -dx11 or -dx12. If your crash is DX12-only, DX11 can avoid it (at some performance cost).
  • Disable overlays: Steam, Discord, GeForce/NVIDIA App, and any recording or RGB software overlays. Each one hooks into Direct3D.
  • Use borderless windowed instead of exclusive fullscreen if the crash happens on alt-tab or when a popup appears.
  • Cap your frame rate with the in-game limiter or RTSS. Uncapped menus running at hundreds of FPS are a common place for this crash to appear.

Fix 5 — The 0x0 ‘S_OK’ variant (MPO)

If your message ends in (Error: 0x0 - 'S_OK'), you are probably seeing a specific, documented case. It was reported in Unreal Engine 4.26 and 4.27 on PCs with both integrated graphics and an NVIDIA card, triggered when an overlay window appeared in fullscreen.

NVIDIA traced this family of problems to multiplane overlay (MPO) behaviour with Game Ready drivers 461.09 and newer, and published a support article with an official file to disable MPO. The simple workaround is to run the game in borderless windowed mode. If you disable MPO, use NVIDIA’s own method rather than a third-party tweak, and note that it is reversible — Windows has changed how the override works in recent Windows 11 builds, so re-check it after a major Windows update.

Fix 6 — Increase the TDR timeout (HUNG only, last resort)

If your code is HUNG, you can lengthen how long Windows waits before resetting the driver by setting the TdrDelay registry value (for example to 8 seconds). This hides the symptom rather than fixing the cause and helps REMOVED far less. Full steps and caveats are in the DEVICE_HUNG guide.

Still crashing?

  • Verify the game files in Steam or the Epic Games Launcher — a corrupted shader or pak file can crash the renderer.
  • Check VRAM: if the crash comes after a long session at high settings, you may be running out of video memory first — see out of video memory error fix and why FPS drops after hours of gameplay.
  • Test another Unreal game. If several UE titles crash the same way, the problem is your PC. If only one does, check that game’s known issues and patches.

The short version

  • The message is a wrapper — the code in brackets is the diagnosis.
  • HUNG = GPU timeout → clean driver, remove overclocks. REMOVED = card lost → overclocks (including XMP/EXPO), power, heat.
  • Try -dx11/-dx12, disable overlays, and use borderless windowed.
  • S_OK on UE 4.26/4.27 with an iGPU + NVIDIA → MPO; borderless windowed or NVIDIA’s MPO fix.

Once the crash is fixed, the rest of your Windows setup still matters for smooth frame times — see the best tools to fix stuttering and our timer resolution guide.

Frequently asked questions

What does 'Unreal Engine is exiting due to D3D device being lost' mean?

It means the game lost contact with your graphics card through Direct3D and Unreal Engine shut down rather than continue without a GPU. The message is a wrapper: the code in brackets after it — usually 0x887A0006 'HUNG' or 0x887A0005 'REMOVED' — tells you the actual cause. It is almost always a driver, overclock, power or heat problem on your PC, not a bug unique to the game.

What does error 0x887A0006 HUNG mean in an Unreal Engine game?

0x887A0006 is DXGI_ERROR_DEVICE_HUNG: the GPU took too long on a piece of work and Windows' Timeout Detection and Recovery (TDR) reset the driver after about two seconds. The usual causes are an unstable GPU overclock, a corrupted driver, or a power or thermal problem. A clean driver reinstall with DDU and removing any overclock fix most cases.

What does 0x887A0005 REMOVED mean in an Unreal Engine crash?

0x887A0005 is DXGI_ERROR_DEVICE_REMOVED: Windows treated the graphics card as if it had been physically unplugged, a more severe loss than a timeout. It points harder at hardware-level instability — power delivery, GPU or memory overclocks including XMP/EXPO, or the card itself. Stock clocks, a clean driver and a check of power connectors are the first steps.

Does switching Unreal Engine games to DirectX 11 fix the D3D device lost crash?

Sometimes. Many Unreal Engine games accept a -dx11 or -dx12 launch option, and switching API can avoid a driver bug that only affects one of them. It is a workaround rather than a fix: if the underlying cause is an overclock, a power problem or heat, the crash will usually follow you to the other API.

Why does the D3D device lost error show 0x0 S_OK?

The 'S_OK' variant was documented in Unreal Engine 4.26 and 4.27 on PCs with both integrated graphics and an NVIDIA card, where it was triggered by overlay windows in fullscreen. NVIDIA traced it to multiplane overlay (MPO) behaviour in Game Ready drivers from 461.09 onward and published a support article with an official way to disable MPO. Running the game in borderless windowed mode is the simpler workaround.