Fix 0x887A0006 DXGI_ERROR_DEVICE_HUNG Game Crashes

That crash usually means a DirectX device got yanked away mid-frame by a driver reset, an overlay, or a bad GPU driver. Here's how to pin down which one and stop it.

Quick answer

0x887A0006 (DXGI_ERROR_DEVICE_HUNG) means the GPU driver stopped responding for more than 2 seconds, Windows reset the device, and your game didn't survive the reset. The real fix is almost always one of three things: an unclean GPU driver install, a third-party overlay hooking into DirectX, or an unstable overclock. Reinstalling Windows won't help — that's the nuclear option people reach for after the first two fixes fail, and it usually fails too.

Why this happens

The Windows Timeout Detection and Recovery system (TDR) watches the GPU. If the driver doesn't post a response within 2 seconds, Windows kills the context and restarts the driver. Games running on DirectX 11 or 12 see that as the device vanishing, and they crash. You'll see the error in the game log, in Event Viewer under Display, or as a flat-out freeze with no dialog box at all.

Common triggers I've seen on real machines: a driver update where the old driver wasn't fully removed (NVIDIA's installer leaves fragments behind, AMD's does too), Discord or MSI Afterburner overlays hooking into the Present() call, and memory overclocks that pass stress tests for 20 minutes but crash a game 90 minutes in. I've also seen it on a Ryzen 7800X3D rig where the culprit was a single stick of DDR5 running EXPO at 6000 CL30 when the memory controller wanted 5600. Nothing to do with the GPU at all.

Fix it, step by step

Check Event Viewer first. Press Win+R, type eventvwr.msc, hit Enter. Go to Windows Logs > System. Look for a Display event (source is usually "Display" or "nvlddmkm"/"amdkmdag"). If you see "Display driver stopped responding and has recovered," that confirms TDR fired. If you see nothing, the crash is happening inside the game and TDR isn't the cause — skip to step 5.
Strip your overlays. Close Discord, GeForce Experience / NVIDIA App overlay, Steam overlay, MSI Afterburner, RivaTuner Statistics Server, Xbox Game Bar, and anything from Razer or Logitech that draws on screen. Launch the game and play for 20 minutes. If it holds, add them back one at a time — the one that kills it is your culprit. In my experience, RTSS and Discord's overlay own this category.
Undo GPU overclocks. MSI Afterburner's default profile isn't "stock" if you ever touched the sliders. Hit the reset button (the circular arrow) and apply. For AMD Adrenalin, go to Performance > Tuning and set it to Default. For NVIDIA, open the NVIDIA App, Tuning section, and disable automatic tuning if it's on. Also disable XMP/EXPO on your RAM in BIOS as a test — if the crash goes away, your memory is the problem, not your GPU.
Clean-install the GPU driver with DDU. Download Display Driver Uninstaller from Wagnardsoft. Boot into Safe Mode (hold Shift while clicking Restart, then Troubleshoot > Advanced Options > Startup Settings > Restart > press 4). Run DDU, pick your vendor (NVIDIA / AMD / Intel), and hit "Clean and restart." Then install the latest driver from the vendor's site — not Windows Update's version, which is always months behind. After the install finishes, you should see your native resolution return and no yellow bang in Device Manager.
Bump TDR delay. If DDU + no overlays still crashes, extend the timeout. Open Registry Editor, go to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers. Right-click in the right pane, New > DWORD (32-bit). Name it TdrDelay, set the decimal value to 10. Create a second DWORD called TdrDdiDelay with the same value. Reboot. This doesn't fix the underlying instability, but it gives the GPU 10 seconds instead of 2 to recover — which is often enough for heavy scenes that were tripping the timer.

If those don't work

Roll the driver back one major version. NVIDIA's 55x series had a string of DXGI crashes on RTX 40-series cards that got fixed two releases later. AMD's 24.x drivers had a similar run on the 7900 XTX. Grab the previous WHQL release from the vendor's archive page and DDU again before installing.

Test with a different game engine. If only Unreal Engine 5 titles crash, it's likely shader compilation hitting a driver bug — try deleting the game's shader cache folder (usually under %LOCALAPPDATA%\<GameName>\Saved\) and let it recompile. If every game crashes, run sfc /scannow and DISM /Online /Cleanup-Image /RestoreHealth from an admin command prompt to rule out corrupted system files.

Last resort: check GPU thermals under load with HWiNFO64. If the hotspot hits 105°C+ on an AMD card or 100°C+ on NVIDIA, the driver is throttling and timing out. That's a repaste job, not a software fix.

Prevention

Stop installing beta or "Game Ready" drivers on day one. Wait a week, read the release thread on Reddit or the vendor forums, and only then update. When you do update, use DDU every time — the 5 extra minutes saves hours of debugging later. And if you're running EXPO or XMP, stress-test with TestMem5 or Karhu for at least an hour before calling the system stable. Games find memory instability that synthetic benchmarks miss.