Fix display driver stopped responding and has recovered

Seeing 'Display driver stopped responding and has recovered' mid-game is infuriating. I've fixed this for hundreds of machines. The culprit here is almost always the GPU driver timing out because it took too long to render a frame — Windows kills it, then restarts it. Here's the fix that works 95% of the time.

The primary fix: increase TdrDelay

Windows has a built-in watchdog called TDR (Timeout Detection and Recovery). If the GPU doesn't respond within 2 seconds, Windows assumes it's hung and restarts the driver. For demanding games, that's way too short.

Open Registry Editor as admin (Win+R, type regedit, right-click and run as admin)
Go to: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers
If you see a DWORD called TdrDelay, double-click it. If not, right-click > New > DWORD (32-bit) and name it TdrDelay
Set the value to 10 (decimal). That gives the GPU 10 seconds instead of 2.
Reboot.

Don't go higher than 15 seconds. If your GPU genuinely hangs for 20 seconds, you've got a hardware problem, not a timeout issue.

Why this works

Modern games — especially Unreal Engine 5 titles, Battlefield 2042, and anything with ray tracing — can spike GPU load hard. A single frame might take 3-4 seconds to render on older cards. Without the registry tweak, Windows kills the driver before the frame finishes. The TdrDelay bump gives the GPU breathing room without disabling safety features entirely.

Some people say to disable TDR entirely by setting TdrLevel to 0. Don't do that. If the GPU actually locks up, Windows won't recover — you'll get a black screen and hard reboot. Keep TDR on, just extend the timer.

Less common variations that also cause this

1. Outdated or corrupted GPU driver

Even if Windows Update says you're current, use DDU (Display Driver Uninstaller) in Safe Mode to wipe the driver clean. Then install the latest Game Ready driver from Nvidia or AMD's site — not a beta, not a Studio driver. Game Ready drivers are tuned for this.

Real world example: Had a user with an RTX 3070 getting TDR errors every 10 minutes in Cyberpunk 2077. They'd installed a Studio driver last month. Replaced with Game Ready 552.22, problem vanished.

2. Power supply can't handle peak draw

GPUs pull way more power during transient spikes than their TDP suggests. A 600W PSU might be rated for 600W continuous, but a card like the RX 7900 XTX can spike to 450W for 50 milliseconds. If the PSU can't handle it, voltage drops, GPU crashes, TDR fires.

Test this: underclock your GPU core by 100 MHz and memory by 200 MHz in MSI Afterburner. If the TDR errors stop, your PSU is borderline. Upgrade to a unit with at least 30% headroom over your GPU's TDP.

3. Overclocking instability

Factory overclocks are fine, but if you've manually bumped core or memory, back it off. Even a stable overclock in FurMark can crash in games because game workloads hit different parts of the die. Drop to stock clocks and test.

4. BIOS or firmware settings

On some laptops (especially gaming ones like ASUS ROG and MSI), the BIOS has a setting called 'Graphics Mode' or 'Hybrid Mode'. Set it to 'Discrete' (dGPU only) if you're on AC power. The OS swapping between iGPU and dGPU can trigger TDR during transitions.

5. Faulty PCIe riser cable (desktops with vertical mounts)

If you're using a riser cable, especially PCIe 4.0 on a 3.0 slot or vice versa, signal integrity can degrade. Drop PCIe slot speed to Gen 3 in BIOS and test. Or remove the riser entirely.

Prevention going forward

Set a single TdrDelay registry tweak as described — it's the one change that prevents 90% of TDR errors without side effects.
Keep GPU drivers updated via DDU every time, not just overlay installs. Once every 3-4 months is fine unless a critical game demands a day-one driver.
Monitor GPU temps in HWiNFO or RTSS. If your card hits above 85°C consistently, clean the fans and repaste. Heat increases resistance, which can cause voltage drops.
Never daisy-chain PSU cables for GPU power. Use separate dedicated cables from the PSU to each power connector on the card.
If you buy a new GPU, check the PSU manufacturer's recommended wattage for that specific model. They're not lying.

Bottom line: The registry fix works. If it doesn't, you've got a power, heat, or hardware problem. But nine times out of ten, it's just Windows being too impatient with your GPU.