The “Display driver stopped responding and has recovered” message, or the screen going black for a few seconds before coming back — that’s a TDR event. Windows detected that the GPU driver hung, waited the TDR timeout, then forcibly reset it. You got lucky that time. If it happens repeatedly, eventually Windows won’t recover cleanly and you’ll get a blue screen instead. For a broader walkthrough, our Complete Guide to Fixing Windows, Browser, and Software Errors is a good next read.
Most of the time this is fixable. Here’s the honest order to try things in.
Update the GPU driver first. Seriously.
The number one cause of display driver crashes is an outdated or buggy driver. Not Windows update’s driver — the actual driver from NVIDIA, AMD, or Intel’s website. Those are newer and better tested on real hardware.
NVIDIA: nvidia.com/drivers → download the latest Game Ready or Studio driver for your GPU. Run the installer, choose Custom → check “Clean installation.”
AMD: amd.com/support → find your GPU → download and install.
Intel (integrated graphics): go to your laptop manufacturer’s support page for the exact model, not Intel’s generic installer — integrated drivers need to be validated for each laptop’s specific configuration.
After the driver update, reproduce whatever triggered the crash (gaming for 30 minutes, watching 4K video, running the same application). If the crash is gone, you’re done. That’s the fix for probably 40–50% of cases.
Temperature. Check it immediately.
GPU overheating causes the exact same TDR crash. The GPU hits its thermal limit, throttles, and when it can’t recover fast enough, the driver crashes. The giveaway: crashes happen during demanding tasks and never during idle. Also: your GPU fan may be running at max speed right before it crashes.
Download HWiNFO64 and watch GPU temperature during the activity that causes crashes. GPU core above 90°C under load is a problem. Above 95°C on most cards is throttling territory — some cards have thermal shutdown around there.
Quick fixes to try: compressed air through the vents (dust is the biggest culprit on older laptops), improving case airflow on desktops, removing the laptop from soft surfaces that block vents. If temperatures are still high after cleaning, the thermal paste between the GPU die and heatsink may need replacing — typically a job for a service center unless you’re comfortable disassembling the machine.
DDU clean reinstall if the update didn’t help
Sometimes updating over an existing driver doesn’t fix things because old driver remnants interfere. DDU (Display Driver Uninstaller) removes everything completely before installing fresh.
- Download DDU from guru3d.com and have the new GPU driver downloaded and ready
- Boot into Safe Mode (hold Shift → Restart → Troubleshoot → Advanced options → Startup Settings → Safe Mode with Networking)
- Run DDU → select your GPU manufacturer → Clean and restart
- After the clean restart (in normal mode), install the fresh driver
This is more involved but worth doing if a standard update didn’t resolve the crashes.
The TDR registry tweak — useful but not a fix
Increasing the TDR timeout gives the GPU more time to recover before Windows considers it crashed. Some people increase it and the crashes stop appearing — but that’s because they’re masking the problem, not fixing it. The GPU is still hanging; Windows is just waiting longer before doing anything about it.
That said, it can stop the blue screens while you troubleshoot: Registry Editor → HKEY_LOCAL_MACHINESYSTEMCurrentControlSetControlGraphicsDrivers → create a new DWORD value called “TdrDelay” → set it to 8 (decimal). Default is 2 seconds. This gives the GPU 8 seconds to recover before a TDR reset.
Do this as a temporary measure while investigating the root cause, not as a permanent solution.
Power supply and overclocking
Two less common causes that are worth considering if nothing else worked:
Overclocking: if the GPU or memory is overclocked — even slightly, even via a profile you set up months ago — reduce everything to stock. Instability under load is one of the most common consequences of aggressive overclocks.
PSU inadequacy on desktops: a power supply that can’t deliver stable voltage on the 12V rail under full GPU load causes exactly these crashes. OCCT’s power supply test monitors voltage during load. The 12V rail dipping below 11.4V consistently under load indicates PSU problems.
When it’s a specific application, not everything
If crashes only happen in one application — a specific game, a specific creative app — it might be that app’s use of DirectX or Vulkan that’s incompatible with the current driver version. Check for application updates first. Then try rolling back the GPU driver one version. NVIDIA and AMD keep archives of all previous drivers.
Also check Event Viewer after a crash: Windows Logs → System → look for “LiveKernelEvent” Event ID 141 around the time of the crash. The details often name the specific component that failed, which narrows down the diagnosis considerably compared to just “the screen went black.”
For the bigger picture when crashes are causing full blue screens, the BSOD guide covers VIDEO_TDR_FAILURE stop codes specifically. And if the GPU is not being detected at all between crashes, see the GPU not detected guide — the DDU reinstall process applies to both. Microsoft’s TDR documentation explains the recovery mechanism in more technical detail if you want to understand what Windows is actually doing during the crash and recovery sequence.
Hardware vs driver: how to tell which is broken
This is actually harder than it sounds. A dying GPU and a buggy driver produce nearly identical symptoms. But there’s a useful test: boot a Linux live USB (Ubuntu works, doesn’t need installation) and run the GPU under load — play a video, run glxgears, or run a graphics test. Linux uses different drivers than Windows. If the crashes stop entirely on Linux, the problem is Windows driver-related. If the GPU still crashes under Linux, the hardware itself is failing.
You don’t need to be a Linux expert for this. Just boot from the USB and stress the GPU for 15–20 minutes. That’s the most reliable hardware/software discrimination test available without replacing components.
Chrome, browsers and WebGL crashes
A specific scenario: driver crashes that happen only when using Chrome or another browser, particularly on pages with video, WebGL, or graphics-intensive web apps. Browsers run their own graphics acceleration stack on top of the GPU driver. Chrome’s hardware acceleration can hit driver bugs that gaming doesn’t trigger, and vice versa.
If crashes are browser-specific: disable hardware acceleration in Chrome (Settings → System → Use hardware acceleration when available → off) or in Firefox (Settings → General → Performance → uncheck Use hardware acceleration). If browser-specific crashes stop, update the GPU driver — browser-compatible driver versions are usually the latest release.
VRAM corruption and stability tests
GPU memory errors cause display corruption and crashes that look like driver failures but are actually hardware degradation. VRAM doesn’t have ECC like system RAM on consumer GPUs, so errors accumulate as the card ages.
MATS (now part of some GPU diagnostic tools) and VRAM-specific stress tests in OCCT can identify memory errors. If you see pixel corruption — random coloured pixels, checkerboard patterns, or visual artifacts specifically during graphics rendering — that’s often VRAM failing rather than a driver issue.
Multiple monitors and driver load
Driving multiple monitors increases GPU driver complexity and the likelihood of driver hangs on certain configurations. If crashes started after adding a second monitor: try with only one display connected. Also check whether the crash correlates with specific multi-monitor operations like switching displays, dragging windows between monitors, or waking from sleep with monitors in an inconsistent state.
Also worth checking: the cable connection quality. Display crashes are sometimes actually HDMI or DisplayPort signal drops that Windows interprets as a driver failure. A marginal cable that drops signal intermittently causes exactly the “screen went black then came back” symptom. Swap the display cable as a quick test.
DirectX and Visual C++ runtime issues
Less commonly, crashes are caused by corrupted DirectX components or missing Visual C++ redistributables that the GPU driver depends on. Run: Settings → Apps → Installed apps → search “Visual C++” → check that Microsoft Visual C++ 2015–2022 Redistributable (both x64 and x86) are installed. If any are missing or show errors: download the latest from Microsoft’s Visual C++ Redistributable downloads page and install them. Then reinstall the GPU driver to ensure it relinks against the corrected runtime.
Windows 11 and driver signing
Windows 11 with Secure Boot and Memory Integrity enabled requires GPU drivers to be digitally signed with Microsoft certification. Drivers that haven’t been updated for Windows 11 requirements may fail this check. Device Manager shows a Code 43 or Code 52 error on the GPU. The fix is updating to a driver version that’s been certified for Windows 11 — NVIDIA, AMD, and Intel have all been issuing updated signed drivers since Windows 11 launched, but very old driver versions from before 2021 may lack the required signatures.
After ruling everything else out
If you’ve done the full driver reinstall with DDU, temperatures are normal, there’s no overclocking, the PSU is adequate, Linux doesn’t crash, and the problem persists — the GPU hardware is likely failing. Consumer GPUs don’t last forever, and under heavy gaming or content creation use, 5–7 years of heavy use is sometimes all you get.
Before buying a replacement: check whether the GPU is still under warranty (NVIDIA GeForce cards typically have 3 years, AMD RX cards vary by brand but often 2–3 years). If in warranty, use the manufacturer’s RMA process. If out of warranty, the repair vs replace math usually favours replacing unless you’re attached to a specific card.
Frequency matters for diagnosis
How often are the crashes happening? Once a month is very different from once an hour. Rare crashes that started recently and are getting more frequent suggest hardware degradation — GPU, thermal paste, or PSU capacitor aging. Crashes that started suddenly at a consistent frequency and haven’t changed pattern since suggest a driver or software trigger — usually something that changed on a specific date (a driver update, a Windows update, an application install).
Use Reliability Monitor (search “reliability history” in Start) to check the exact date the crashes started. This often immediately reveals whether an update or installation happened that same day, giving you a clear direction for the fix.
The “night-before” trick for GPU driver issues
If you need to track down which driver version caused a problem: NVIDIA keeps a full driver archive at nvidia.com/drivers → All versions. AMD keeps one at amd.com/drivers → Previous Drivers. You can go back specific versions to find the last stable driver before the crashes started.
Look at the date in Reliability Monitor, compare it to when a driver update was installed, and download the driver that was active the day before the crashes started. That’s usually stable. Reinstall it via DDU and set Windows to not automatically update the GPU driver: Device Manager → right-click GPU → Properties → Driver → Update Driver → Search automatically settings → Turn off updates for this specific driver. You’ll get a notification prompt before any future driver updates rather than them installing silently.
Non-GPU causes that look like GPU crashes
Last point worth making: not all “display driver stopped responding” events are actually GPU driver problems. RAM failures and CPU instability can cause graphics subsystem crashes that Windows attributes to the display driver. If the driver update, DDU reinstall, and temperature checks all check out fine, run Windows Memory Diagnostic or MemTest86 to rule out RAM. A single RAM error can cause GPU driver TDR events with no other obvious symptoms.
Similarly, system instability from an overcrowded SATA controller or a failing storage drive occasionally causes enough general system instability to trigger GPU driver timeouts. Running a CHKDSK on the system drive and checking drive health in CrystalDiskInfo covers this base.
One practical thing before ending: take a screenshot or photo of the exact error message when it appears (the recovery notification at the top right of the screen). The specific driver name mentioned in that notification tells you which display adapter crashed — useful on systems with both integrated and discrete graphics, where the crash might be on the integrated GPU rather than the dedicated one, pointing you to completely different driver files and different fix paths. Our guide on Windows 11 Audio Not Working covers an adjacent issue.
And if you’re seeing the crash notification but your work isn’t being lost and the computer keeps running — appreciate that Windows is handling it better than it used to. TDR recovery is a feature, not a failure. A crash that recovers gracefully and lets you keep working beats the alternative. But still fix it, because recovery doesn’t always succeed. See also File Explorer Not Responding in Windows 11 for a related case.
