Fix 0x000000EA Thread Stuck in Device Driver Loop

You launched a game, the screen froze, and now Windows bluescreens with 0x000000EA every time you boot. Annoying. Let's fix it.

The fix

Boot into Safe Mode, rip out the current graphics driver with DDU, then install a fresh driver offline. That's the whole job.

Force Safe Mode. If Windows won't reach the desktop, hold the power button until it shuts off mid-boot. Do that twice. On the third boot you'll get the Recovery Environment. Choose Troubleshoot → Advanced options → Startup Settings → Restart, then press 4 for Safe Mode.
Download DDU before you need it. If you can reach a desktop, grab Display Driver Uninstaller from Wagnardsoft now and drop it on the desktop. If you can't, use another PC and a USB stick.
Run DDU in Safe Mode. Launch it, pick GPU in the device type dropdown, pick your vendor (NVIDIA, AMD, or Intel), then click Clean and restart. Don't touch the "Prevent downloads from Windows Update" checkbox yet.
Block Windows Update from pushing a driver. Back in normal Windows, before you install anything, run this in an admin Command Prompt:
```
reg add "HKLM\SOFTWARE\Microsoft\Windows\CurrentVersion\DriverSearching" /v SearchOrderConfig /t REG_DWORD /d 0 /f
```
That stops Windows from silently reinstalling a broken driver the moment you hit the desktop.
Install a known-good driver. Not the newest. Go back one or two versions. For NVIDIA, the Studio branch is often more stable than the Game Ready branch when a specific title is crashing. Install, reboot, launch the game.

If you can't even stay in Safe Mode long enough to run DDU, you've got a hardware failure or corrupt system files, and no driver trick will save you. Skip to the variations section below.

Why this works

The stop code 0x000000EA literally means THREAD_STUCK_IN_DEVICE_DRIVER. The GPU driver has a worker thread that's supposed to finish a task — draw a frame, flush the command buffer, whatever — within a timeout window. Windows gives it about 2 seconds by default. If the thread doesn't come back, the kernel panics and bluescreens instead of hanging forever.

What's actually happening here is that the driver got into a state where its internal command queue is deadlocked. Often that's an incomplete driver install, a leftover .sys file from a previous version, or a driver that hates a specific DirectX 12 call your game makes. The driver itself is fine on disk. It's the loaded state in kernel memory that's poisoned.

The reason step 3 works is that DDU deletes the driver files, the registry entries under HKLM\SYSTEM\CurrentControlSet\Services\nvlddmkm (or amdkmdag for AMD), and the leftover C:\Windows\System32\DriverStore\FileRepository packages. A normal uninstall from Device Manager only removes the top layer. The stuck kernel thread restarts clean next boot.

And the reason step 5 matters: newer isn't better. Sometimes a driver version has a regression that only shows up on specific GPU architectures — a 0x000000EA that hits RTX 3080 owners and not RTX 4080 owners, for example. Rolling back one version sidesteps the bug while the vendor patches it.

Less common variations

BSOD happens at idle, not in a game

If you crash while just sitting on the desktop, the culprit is usually a background process touching the GPU — hardware acceleration in Discord, Chrome, or a video player. Disable hardware acceleration in each app and see if it stops. If it does, you've isolated the caller. Now update that app or leave acceleration off.

Laptop with switchable graphics

Intel iGPU + NVIDIA dGPU laptops can hit 0x000000EA when the driver handoff between the two fails. The fix is different: grab both the Intel and NVIDIA drivers from the laptop manufacturer's support page, not from Intel.com or NVIDIA.com. OEM drivers include the handoff firmware that generic drivers skip. Install Intel first, reboot, then NVIDIA.

TDR timeout is too short for your workload

Some pro apps and a few badly-optimized games need more than 2 seconds. Bump the timeout without disabling TDR entirely:

reg add "HKLM\SYSTEM\CurrentControlSet\Control\GraphicsDrivers" /v TdrDelay /t REG_DWORD /d 10 /f
reg add "HKLM\SYSTEM\CurrentControlSet\Control\GraphicsDrivers" /v TdrDdiDelay /t REG_DWORD /d 10 /f

Reboot. This buys your GPU ten seconds instead of two. Don't set TdrLevel to 0 — that turns crash detection off and you'll get silent freezes instead of a diagnostic bluescreen. Worse trade.

Overclock or undervolt got unstable

If you're running MSI Afterburner or an undervolt profile, reset to stock before doing anything else. A too-aggressive undervolt can cause the GPU to hang mid-frame, which surfaces as 0x000000EA. Stock clocks first. Then reapply the profile in smaller steps.

Prevention

Use DDU for every driver upgrade, not just problem fixes. It takes four minutes and saves an afternoon.
Keep the DriverSearching registry value at 0 permanently. Windows Update will still deliver security patches but won't shove a GPU driver on you at 3 AM.
Before installing a new driver for a game launch, check the game's subreddit or Steam forums for that version number. If three people say it crashed their 4070, wait a week.
If you use an undervolt, benchmark it with a proper stress tool for 30 minutes before trusting it in a game. Crashes that show up at minute 40 aren't driver bugs — they're your profile.

One more thing: if DDU and a rollback don't clear the loop, and you're seeing it on a card more than four years old, suspect the card. VRAM decay and failing power delivery produce identical symptoms. Test it in another machine before you spend more hours chasing drivers.