DXGI_ERROR_DEVICE_REMOVED 0x887A0005 | Stop The Crashes

If a game, Unreal project, or render tool keeps throwing this error, you’re not alone. The tricky part is that the message sounds like a dead GPU, yet most cases come down to the driver timing out, a power dip, a bad overlay, or a setting that spikes load at the wrong moment.

This guide walks you through fixes in the same order I’d use on a real PC: start with fast checks that don’t touch the registry, then move into driver cleanup, stability tuning, and only then the heavier steps.

DXGI_ERROR_DEVICE_REMOVED 0x887A0005

Windows has a safety feature called Timeout Detection and Recovery (TDR). When the GPU or its driver stops responding for long enough, Windows resets the driver so the desktop can keep running. Microsoft documents this TDR flow for Windows Display Driver Model drivers, including how a timeout is detected and how the system recovers.

From the app’s point of view, that reset looks like the graphics device vanished. DirectX then returns a “device removed” reason such as 0x887A0005, and the game or tool shuts down to avoid corrupt frames or a hang.

You’ll often see the error after one of these moments: a sudden FPS spike in a menu, a shader compilation burst, an alt-tab, a capture overlay hooking the frame pipeline, or a quick power swing when the GPU boosts clocks.

What the code is telling you

• Driver reset happened — Windows restarted the display driver during a GPU task, so the app lost its rendering context.
• Something triggered the reset — load spikes, unstable clocks, heat, power limits, buggy drivers, or a bad hook can all push the driver over the edge.
• The fix is often stability — not a single magic tweak, but a short chain of steps that removes triggers and steadies the GPU.

Fast triage that saves time

Before you change anything big, try these quick checks. They’ll tell you whether you’re chasing a system-wide stability issue or a single game setting.

Confirm the pattern

• Test a second game — run another DirectX title for 10–15 minutes; if both crash, aim at drivers, power, clocks, or heat.
• Note the moment — if it happens on alt-tab, fullscreen switching, or opening an overlay, start with display mode and overlays.
• Check Event Viewer — look for Display driver events around the crash time; a driver reset points straight at TDR behavior.

Drop the load spike right away

• Cap frame rate — set an in-game FPS limit or use your driver’s limiter; menus can hit 500+ FPS and cause sharp power swings.
• Lower one heavy setting — reduce ray tracing or shadows first, then retest; the goal is a calmer peak, not a blurry game.
• Switch to borderless — borderless windowed can avoid mode-switch churn on some setups.

Turn off common hook points

• Disable overlays — turn off GeForce Experience overlay, Discord overlay, Steam overlay, Xbox Game Bar, and screen recorders, then retry.
• Unplug extra monitors — run one display for a test session; mixed refresh rates can add instability on some driver stacks.
• Reset GPU tuning — return MSI Afterburner, Adrenalin tuning, or BIOS OC to stock for the test.

Fixing dxgi error device removed 0x887a0005 on Windows 10 and 11

If the quick steps helped, you’ve learned the crash is trigger-sensitive. Now you want a stable baseline that stays stable after updates. The steps below center on Windows 10 and Windows 11 because both use the same TDR model, and the same driver stack ideas apply.

Start with Windows and chipset updates

• Install Windows updates — open Settings, run Windows Update, and install optional driver and platform updates when offered.
• Update chipset drivers — grab the latest chipset package from your motherboard or laptop maker, not a random mirror.
• Update BIOS if needed — if your BIOS is old, a newer build can fix PCIe, memory training, or power behavior.

Set Windows graphics and power options

These settings don’t fix broken drivers, yet they can reduce the spikes that trigger a reset.

• Use a stable power plan — set Windows to Balanced or High performance, then keep CPU minimum state sane, not pinned at 100%.
• Toggle Hardware-accelerated GPU scheduling — switch it on or off in Windows Graphics settings, reboot, then test the same scene.
• Disable PCIe link power saving — in Advanced power settings, set Link State Power Management to Off for a test run.

Do a clean GPU driver install

A “clean install” is more than clicking update in a launcher. You want to remove old shader caches, leftover profiles, and mismatched components.

• Download one driver version — pick the newest stable release from NVIDIA, AMD, or Intel, or use a known-good older version if the crash started after an update.
• Remove the current driver cleanly — use Display Driver Uninstaller (DDU) in Safe Mode when possible, then reboot.
• Install with minimal extras — skip overlays and recording tools during setup, then add them back only if you need them.

Pick the right graphics API and mode

• Try DirectX 11 — if a game offers DX11 and DX12, test DX11 first; DX12 can expose driver edge cases on some rigs.
• Disable “true fullscreen” — if the crash hits during alt-tab, run borderless to reduce mode swaps.
• Clear shader cache — after driver changes, clear shader caches in the driver panel or delete the game’s shader cache folder, then let it rebuild.

Stability checks that catch the real culprit

At this point, the crash is often tied to stability under peak load. That can be heat, power delivery, VRAM errors, unstable RAM, or a boost curve that’s fine in benchmarks yet flaky in games.

Watch temperatures and clocks during a crash window

• Log sensor data — use HWiNFO or GPU-Z logging to capture GPU temp, hotspot, VRAM temp, GPU power, and clock boosts.
• Check hotspot gaps — a high hotspot-to-core delta can point to paste or mounting issues on some cards.
• Verify fan behavior — confirm fans ramp under load; silent fan curves can cook a card without sounding loud.

Remove instability sources one by one

• Undo GPU overclocks — reset core and memory to stock; a small memory OC is a common trigger for DXGI device removal.
• Undo undervolts for testing — undervolts can be stable in one title and crash in another; test stock first, then tune again.
• Set a mild power limit — on some cards, a small power limit reduction smooths spikes and stops resets.

Check the boring hardware basics

• Reseat the GPU — fully power down, remove the card, reseat it, and reconnect PCIe power cables until they click.
• Use separate PCIe cables — don’t daisy-chain one cable to two 8-pin plugs when the PSU includes multiple leads.
• Test memory stability — turn off XMP/EXPO for a test run; unstable system RAM can crash the graphics stack.

When TDR tuning helps and when it hurts

Sometimes the GPU is fine, and the workload is just heavy enough that Windows decides it’s hung. Microsoft’s TDR documentation explains that Windows detects a GPU timeout and recovers by resetting the driver. Microsoft also publishes registry entries used for TDR testing and debugging. Third-party apps like Adobe Substance 3D Painter document TdrDelay as a way to give long GPU tasks more time.

This is not a first step for games. A higher delay can mask a crash, and it can also turn a short hiccup into a long freeze. Use it when your issue is long compute tasks, heavy shader compilation, or creator workloads that are known to trip TDR.

Safe rules before you touch the registry

• Create a restore point — System Protection lets you roll back if you mis-type a value.
• Change one value only — don’t stack multiple tweaks, or you’ll lose the trail.
• Use small steps — start modestly; a giant number can leave you staring at a frozen screen.

Add or adjust TdrDelay

• Open Registry Editor — press Win+R, type regedit, then run as admin.
• Go to GraphicsDrivers — browse to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers.
• Create TdrDelay — add a DWORD (32-bit) value named TdrDelay if it’s missing, then set it in seconds.
• Reboot and retest — restart Windows and rerun the exact workload that triggers the crash.

Undo the change if games start freezing

• Remove TdrDelay — delete the value or set it back to default behavior, then reboot.
• Fix the root trigger — return to load spikes, overlays, clocks, temps, and driver cleanup steps.

A repeatable checklist to stop DXGI crashes

If you want one flow you can follow every time, use this order. It keeps you from bouncing between random “fixes” and losing track of what helped.

Session one: quick stability pass

1. Cap FPS — set an in-game limit and match it to your monitor’s refresh for the test.
2. Disable overlays — turn off capture and chat overlays that hook the render path.
3. Run stock clocks — reset GPU and RAM tuning to default and retry the same scene.
4. Switch API — try DX11 if DX12 is crashing, then keep the one that stays stable.

Session two: clean baseline

1. Update Windows — apply all updates and reboot.
2. Update chipset — install current platform drivers from the system maker.
3. DDU reinstall GPU driver — wipe the driver in Safe Mode, then install one known-good version.
4. Clear shader caches — delete old caches so the new driver rebuilds them cleanly.

Session three: hardware sanity checks

1. Check temps under load — log hotspot, VRAM temp, and power draw while reproducing the crash.
2. Inspect power cabling — use separate PCIe leads and confirm connectors are fully seated.
3. Test RAM stability — disable XMP/EXPO for a run, then re-enable only if the crash stays gone.

If the error still shows up after all three sessions, the odds tilt toward a hardware fault: a PSU that dips under transient load, a GPU with unstable VRAM, or a motherboard slot issue. At that stage, swap-testing one part at a time beats endless setting tweaks.

When you see the message “dxgi_error_device_removed 0x887a0005” again, treat it as a clue, not a verdict. Most systems recover with clean drivers, calmer load spikes, and stock-stable clocks.

References: Microsoft TDR overview, Adobe notes on TdrDelay.

Send us your feedbackTell us what you thought about this post

Name * Email *

Your comment *

Quick check — what is …? *

Send feedback