You're 40 minutes into a raid, the screen freezes for two seconds, then everything drops to the desktop. No BSOD. No warning. Just you and a black screen. That's the classic GPU driver error signature, and it shows up under a dozen different names: Xbox Game Bar says nothing, Windows Event Viewer logs a TDR event, and the game just closes.
This flow goes from a 30-second fix to a 15-minute deep dive. Try each step in order and stop the moment your games run clean. Most people never make it past step 2.
Step 1: The 30-Second Fix (Try This First)
Before you touch drivers, rule out the dumb stuff. A stale process holding the GPU hostage causes more "driver crashes" than actual driver crashes.
- Save everything and close your game completely. Not minimized. Fully quit.
- Press
Ctrl + Shift + Escto open Task Manager. - Look for anything running in the background: Discord overlay, MSI Afterburner, RivaTuner, OBS, GeForce Experience overlay, Razer Synapse. Kill them all. Right-click each one, End Task.
- Reboot. Not shut down. Reboot. Fast Startup on Windows 10 and 11 doesn't fully reload the GPU driver on a shut down, and that catches people out constantly.
- Launch the game and play for 10 minutes. If it holds, you had a conflicting overlay. Add your overlays back one at a time until you find the culprit.
After the reboot, you should see the game load normally and hold framerate. If it crashed in the same spot again, move on.
Quick check: is it one game or all games?
If only one title crashes, the fix is probably the game, not your driver. Verify the game files through Steam (right-click game > Properties > Installed Files > Verify integrity), or through Epic (three dots > Manage > Verify). A corrupted shader cache can look exactly like a GPU driver fault. This is especially common after a game update drops mid-week, like when Warzone pushes a hotfix and your old shader cache no longer matches.
Step 2: The 5-Minute Fix (Clean Driver Reinstall)
If every game crashes, you've got a driver problem. And here's the opinionated part: a standard "Express Install" won't fix it. You need a clean install, and the cleanest way is DDU.
Download Display Driver Uninstaller from Wagnardsoft. Yes, it's a third-party tool. Yes, it's the one every technician uses. Grab the current driver from NVIDIA, AMD, or Intel first so you're not stuck with no display driver after the wipe.
- Disconnect your ethernet cable or turn off Wi-Fi. Windows Update will try to shove a generic driver back in the middle of your reinstall. It loves doing that.
- Boot into Safe Mode. Hold
Shiftwhile clicking Restart, then Troubleshoot > Advanced Options > Startup Settings > Restart > press 4 for Safe Mode. - Run DDU. Pick your GPU vendor, click Clean and restart. The screen will flicker and go dark for a few seconds. That's normal. Don't panic.
- Once you're back at the desktop in normal mode, run the driver installer you downloaded. Choose Custom Install and check Perform a clean installation.
- During install, skip GeForce Experience (or Adrenalin's Streaming tab if you don't use it). Fewer hooks into your display stack means fewer things to break.
- Reboot one more time. Reconnect the network.
After the reboot, you should see your GPU correctly named in Device Manager (right-click Start > Device Manager > Display adapters), with no yellow warning triangle. Launch your game and play for 30 minutes. If it's solid, you're done. If it still crashes, the driver wasn't your problem — the hardware or the OS is.
Version-specific notes
NVIDIA's 55x.xx branches had documented TDR issues on RTX 30-series cards with multi-monitor setups. AMD's Adrenalin 24.x line had a stretch where driver timeouts spiked on 6000-series cards with hardware acceleration enabled in Chrome. If you're on either, roll back one branch and retest.
Step 3: The 15+ Minute Fix (Dig into the Logs)
Still crashing? Now we get forensic. Open Event Viewer (eventvwr.msc) and look in Windows Logs > System. Filter for these sources: Display, nvlddmkm, amdwddmg, WHEA-Logger, and BugCheck.
| Event ID | Source | What it usually means |
|---|---|---|
| 4101 | Display | TDR timeout — driver stopped responding and recovered |
| 13 / 14 | nvlddmkm | NVIDIA driver crashed the GPU context |
| 1 | WHEA-Logger | Hardware-level fault, often PCIe or VRAM |
| 1001 | BugCheck | Blue screen with a stop code |
Event 4101 is the smoking gun for "display driver stopped responding and has recovered." It means Windows waited 2 seconds (default TDR delay), the GPU didn't answer, and Windows reset it. The screen flicker you saw is that reset.
Bump the TDR delay
Sometimes the fix isn't a crash — it's an aggressive timeout. Heavy scene loads in games like Cyberpunk 2077 with path tracing on can trip the 2-second limit even when nothing's wrong. Bump it:
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 /fReboot after. You should see the same game load without the two-second freeze-and-recover cycle. This masks the symptom; it doesn't fix a real hardware fault. If you're seeing WHEA-Logger events alongside the display errors, the TDR tweak won't save you — that's the GPU telling you it's failing.
When it's actually hardware
If you've done a clean driver install on a known-good branch and you're still getting WHEA errors, run a stress test. OCCT's VRAM test for 15 minutes or FurMark for 10 will surface it. Artifacts, black screens during the test, or WHEA entries in the log mean the card is dying. RMA it. No driver on earth fixes a failing memory chip.
Sanity checks before you blame the GPU
- Power supply wattage: an RTX 4070 Ti pulls 285W under load plus transient spikes. A 550W no-name PSU causes driver crashes that look identical to a bad card.
- PCIe seating: pull the card, blow out the slot, reseat it. Loose seating throws WHEA errors that vanish forever after a reseat.
- Motherboard BIOS: PCIe 4.0 stability bugs shipped on several B550 and X570 boards. Update the BIOS before you RMA anything.
- Monitor cable: a cheap DisplayPort cable at 4K/144Hz causes signal drops the driver reports as timeouts. Swap it with the one that came in the box.
Work through it in order. Thirty seconds, five minutes, fifteen minutes. The vast majority of "GPU driver errors" are fixed before you ever open a log file — but when they aren't, the log tells you exactly which wall you're hitting.