GPU Driver Crashes in Games: Fix the Black Screen & TDR Error

That popup — "Display driver nvlddmkm stopped responding and has successfully recovered" — is a special kind of infuriating. The game freezes, the screen goes black for two seconds, and you're back at the desktop with a dead framerate and lost progress. I've dealt with this on everything from a GTX 970 to a brand-new RX 7800 XT. The error isn't the driver dying. It's Windows giving up on it.

Here's the breakdown of what's actually happening and the fixes that work, starting with the one that solves it for most people.

Cause 1: A Bad Driver Install (The #1 Culprit)

Most TDR errors trace back to a corrupted or half-installed driver. This happens when Windows Update pushes a driver over your existing one, or when you install a new GPU driver without removing the old one. The leftover registry entries and DLLs from your previous driver fight with the new one, and the GPU eventually throws a timeout error.

The fix is a clean install using Display Driver Uninstaller (DDU). Do not just uninstall from Device Manager — that leaves files behind.

1. Download DDU from Wagnardsoft's site and extract it to a folder.
2. Boot into Safe Mode. On Windows 11, hold Shift while clicking Restart, then choose Troubleshoot > Advanced Options > Startup Settings > Restart and press 4.
3. Run DDU. Select your GPU vendor (NVIDIA, AMD, or Intel) and click Clean and restart.
4. After the reboot, install the latest driver from your GPU maker's site. Skip the driver Windows Update tries to install.

If you're on an NVIDIA card and this is happening with a brand-new driver, roll back one version. NVIDIA's 545.xx series had a nasty TDR regression on RTX 3000 cards in games like Cyberpunk 2077. Sometimes the latest isn't the greatest.

Cause 2: Overheating or an Unstable Overclock

Your GPU doesn't like being hot, and it really doesn't like unstable clocks. When the GPU hits a thermal limit or an overclock that can't sustain a load, it stalls long enough for Windows to trigger TDR.

Check your temps first. Anything above 85°C under load is a problem. Use HWiNFO64 or MSI Afterburner to log temps during a game session.

If you've overclocked — even a factory overclock — dial it back. Open MSI Afterburner and reduce the Core Clock offset by -50 MHz. If that doesn't help, try -100. You can also set a custom fan curve to ramp up earlier. A stable 70°C beats an unstable 80°C every time.

I've seen more "faulty GPUs" turn out to be bad factory overclocks than actual hardware failures. Try the underclock before you RMA anything.

Cause 3: TdrDelay Is Set Too Low

Windows has a built-in watchdog that resets the GPU if it doesn't respond within 2 seconds. That's fine for desktop use, but heavy games can legitimately need more time. Bumping the TdrDelay value gives your GPU breathing room.

Open Registry Editor and navigate to:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers

Create a new DWORD (32-bit) Value called TdrDelay and set it to 8 (decimal). If you're on a heavy modded game, 10 is fine. Reboot and test.

You can also set TdrDdiDelay to the same value for good measure. If this makes the crashes stop but they come back later, the underlying cause is probably still there — this is a bandage, not a cure.

Less Common: Power Supply or PCIe Issues

If none of the above works, check your power supply. A failing PSU can deliver inconsistent voltage under load, and the GPU will glitch. Reseat the GPU in its PCIe slot and try a different 8-pin cable. On older systems, make sure the PSU wattage actually covers your card's peak draw.

Quick Reference Table

Start with the DDU clean install. If that doesn't do it, watch your temps. The registry tweak is a last resort before you start swapping hardware. And if you're still getting TDR crashes after all of this, your card might genuinely be dying — but you'll have ruled out the easy stuff first.