Fix 'Display driver stopped responding and has recovered'

This error means your graphics driver crashed and Windows reset it. We'll go from a 30-second driver restart up to a full clean reinstall and registry TDR tweaks.

What that error actually means

You're 20 minutes into a ranked match, the screen freezes, goes black for two seconds, and when it comes back you see a little balloon: "Display driver stopped responding and has recovered." Sometimes the game survives. Sometimes it doesn't, and you're staring at a desktop.

That bubble is Windows' Timeout Detection and Recovery (TDR) kicking in. Your GPU stopped answering within a 2-second window, so Windows killed and restarted the driver before it took the whole system down. This is almost never an actual hardware failure on the first occurrence. It's a driver bug, an unstable overclock, a power delivery hiccup, or a game doing something stupid with the GPU command queue.

Before you start: check Event Viewer > Windows Logs > System for an event ID 4101 from source Display. That confirms it's a TDR crash and not something else masquerading as one. If you also see event 41 (Kernel-Power) alongside it, your problem might be the PSU, not the driver. Keep that in mind as we go.

Fix 1: The 30-second reset

Try this first. It's dumb, but it clears the current stuck state and works more often than you'd think.

Press Windows + Ctrl + Shift + B at the same time.
Your screen will flash black for a second and you'll hear a beep.
The GPU driver has now been forcibly restarted.

What to expect: the flash and beep are normal. If your game window recovered on its own, great, but the underlying cause is still there and it will likely happen again. This is a bandage, not a fix.

If you're overclocking the GPU with MSI Afterburner, EVGA Precision, or the AMD Adrenalin tuning panel, reset everything to stock right now. Same for GPU memory. An unstable core clock is the number one cause of TDR on cards that otherwise work fine. Hit the reset button in your tuning tool, apply, then reboot. If the crash stops, there's your answer.

Fix 2: The 5-minute driver clean reinstall

If you've been installing "Game Ready" drivers on top of each other for two years, your driver store is a mess. A clean install fixes a huge chunk of TDR reports.

Download DDU (Display Driver Uninstaller) from Guru3D. Get the latest version, not whatever mirrors Google coughs up.
Download the current driver for your card from NVIDIA, AMD, or Intel. Save the installer but don't run it yet.
Disconnect from the internet. Seriously. Windows Update will try to shove an older driver on you the moment DDU finishes.
Boot into Safe Mode. Hold Shift while clicking Restart from the Start menu, then choose Troubleshoot > Advanced Options > Startup Settings > Restart, and press 4.
Run DDU. Pick your GPU vendor from the dropdown. Click Clean and restart.

After it reboots, install the driver you downloaded. During NVIDIA setup, choose Custom (Advanced) and tick Perform a clean installation. For AMD, use the Factory Reset option in the installer.

What to expect: the first boot after DDU will look weird. 800x600 resolution, no hardware acceleration. That's fine and temporary. Once the new driver installs, everything goes back to normal.

Now launch the game that was crashing. Play for at least 30 minutes. If it's stable, you're done. If not, keep reading.

Fix 3: The advanced work (15+ minutes)

3a. Check your power supply and cables

Here's the part most guides skip. TDR spikes during big GPU load changes are frequently a PSU problem. If your 3080 is on a 550W unit that's five years old, you're right at the edge. When the card pulls a transient spike, voltage sags, the driver times out, and Windows recovers it.

Test: run the game with a negative power limit. In MSI Afterburner, slide the power limit to 70% and apply. Play for 30 minutes. If the error is gone, your PSU is likely the culprit. You need more headroom. Aim for 1.5x your GPU's TDP plus CPU load.

While you're in the case: reseat the PCIe power cables. Use two separate cables from the PSU if your card has two connectors, not one daisy-chained cable. This alone fixes some 30-series crashes.

3b. Bump the TDR delay in the registry

This doesn't fix an unstable card. What it does is give the GPU more time to respond before Windows pulls the plug. Some games (looking at you, driver-heavy simulation titles and modded Skyrim) legitimately need more than 2 seconds during heavy shader compilation.

Open regedit and navigate to:

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers

Right-click in the right pane, New > DWORD (32-bit) Value. Name it TdrDelay. Double-click and set it to 10 (decimal). That's 10 seconds instead of 2.

Optionally, add TdrDdiDelay with the same value if you're still getting crashes during launch.

What to expect: reboot after this change. If the game was crashing at a specific demanding moment, it now has more time to work through it. If the GPU was genuinely dead or unstable, it just freezes for 10 seconds instead of 2 before doing the same thing. Set these back to default (delete the keys) once you've confirmed the fix.

3c. Roll back or roll forward the driver

Not every new driver is better. NVIDIA's 545-series, for example, had well-documented TDR problems on 20-series cards. AMD has had similar rough launches.

In Device Manager, right-click your GPU, Properties, Driver tab. Roll Back Driver if it's greyed out, Windows doesn't have an old one cached, so you'll need to grab an older version manually from the vendor's archive.

Rule of thumb: if the crash started right after a driver update, roll back two versions and test. If you're on a driver older than six months, update to current and test. Don't guess, change one variable at a time.

3d. Test the card cold

If the crash only happens after 20+ minutes of play, heat is involved. Install HWiNFO64 and watch your GPU hotspot temperature during a session. If the hotspot is hitting 105°C or higher while the core reads 75°C, your thermal pads or paste have dried out. That's a repaste job.

A quick sanity check: take the side panel off, aim a desk fan into the case, and play. If the crash disappears, you've confirmed a thermal issue. The fix is new thermal paste and pads, not a driver tweak.

When to stop and RMA

If the card throws a TDR on a fresh Windows install with no overclock, a known-good PSU, and stock clocks in a well-ventilated case, you're looking at hardware failure. VRAM degradation and dying GPU cores both present exactly like this. Test in another machine if you can. If it fails there too, open an RMA ticket with the manufacturer.

One last thing: don't run Furmark for hours "to test stability." It doesn't reflect real gaming loads and can damage cards that are already marginal. Use your actual game, at your actual settings, for the actual amount of time you normally play. That's the only test that matters.