The fix that works for most people
If you’re staring at a blue screen that reads VIDEO_TDR_FAILURE (nvlddmkm.sys), do this first: fully remove your NVIDIA driver in Safe Mode with a cleanup tool, then install a fresh copy. Not “Update driver.” A clean wipe. That single move resolves the majority of these crashes, because the file named in the crash is your NVIDIA display driver, and the problem is almost always the driver stack, not Windows itself.
Here’s the short version, and then I’ll walk it through properly. Download the latest NVIDIA Game Ready or Studio driver ahead of time. Grab Display Driver Uninstaller (DDU), the free tool from Wagnardsoft that the repair community leans on. Boot into Safe Mode. Run DDU, pick “Clean and restart.” Install the driver you downloaded. Done for most people.
The clean reinstall, step by step
This is the one that matters, so here are exact clicks. Works on Windows 10 and Windows 11, with the divergences flagged.
- Download the driver first, while your PC still boots. Go to nvidia.com/drivers, enter your GPU model, and download the current Game Ready driver. Save it somewhere you’ll find it. Do this before you remove anything, so you’re not stranded with no display driver.
- Download DDU. Get Display Driver Uninstaller from wagnardsoft.com. Extract the zip to your desktop. Don’t run it yet.
- Boot into Safe Mode. On Windows 11 and 10: Settings > System > Recovery > Advanced startup > Restart now. Then Troubleshoot > Advanced options > Startup Settings > Restart, and press 4 (or F4) for Safe Mode. DDU strongly recommends running in Safe Mode so no NVIDIA components are loaded and locked.
- Run DDU. Open Display Driver Uninstaller.exe. On the right, select device type GPU and vendor NVIDIA. Click Clean and restart. It strips the driver plus leftover registry entries and folders, the stuff a normal uninstall leaves behind.
- Install the fresh driver. After the reboot, run the NVIDIA installer you downloaded in step 1. Choose Custom (Advanced) and tick “Perform a clean installation.” Reboot when it finishes.
How to confirm it worked: open Device Manager, expand Display adapters, right-click your NVIDIA GPU and check that it reports “This device is working properly.” Then actually stress it. Play the game or run the workload that triggered the crash for 20 to 30 minutes. No blue screen under load is your real pass.
When the newest driver is the problem
Here’s the counterintuitive part people miss. Updating doesn’t always fix this. Sometimes the latest driver is the unstable one, and the crash started the day you installed it or the day Windows Update pushed a new version behind your back.
If your nvlddmkm.sys crashes began right after a driver update, roll back. In Device Manager, right-click your GPU, choose Properties > Driver tab > Roll Back Driver. If that button is greyed out, use DDU again and install a slightly older driver from NVIDIA’s download archive instead of the current one. A known-stable release from a month or two back beats a shiny broken one every time.
Also check Windows Update, including optional updates (Settings > Windows Update > Advanced options > Optional updates). Windows sometimes installs its own NVIDIA driver that conflicts with yours. If you see a display driver sitting in optional updates, that’s a suspect worth eliminating.
One real case, and what finally stopped it
Take a fairly typical setup: an RTX 3070 in a Windows 11 machine that ran clean for months, then started throwing VIDEO_TDR_FAILURE only during gaming, never at the desktop. That “only under load” pattern is a tell. It points at the GPU being pushed, not at Windows file corruption.
The clean DDU reinstall helped for a day, then the crashes came back. So the next move was to strip the overclock. This card had a mild factory-ish overclock applied in MSI Afterburner. Setting the core and memory clocks back to stock, and clearing any undervolt, stopped the crashes cold. Microsoft support threads and repair guides both flag overclocking, shaky power delivery, and thermal stress as things that intensify this exact error, and testing at stock settings is how you rule them out.
The lesson from that machine: a driver reinstall that fixes the crash for a few hours and then fails again is telling you the driver was never the whole story. Something is pushing the GPU past stable. Back everything to stock, retest, and only then decide whether the driver or the hardware is at fault.
Bad advice to skip, and the causes worth knowing
A few widely-shared “fixes” for the nvlddmkm.sys blue screen error waste your time or your money.
- “Rename nvlddmkm.sys and copy a fresh one from the driver folder.” This old trick floats around every forum. It’s fragile, it breaks on modern driver packages, and a proper DDU reinstall replaces that file cleanly anyway. Skip it.
- “Run a registry cleaner” or paid one-click repair apps. Registry cleaners don’t fix graphics driver crashes and can cause new problems. There’s no magic button here.
- “It’s a virus, delete it.” nvlddmkm.sys is a legitimate NVIDIA kernel-mode driver file. Deleting it just leaves you with no display driver. It’s the name Windows reports in the crash, not malware.
That last point is the big misconception. The file named in the blue screen is the symptom, not the root cause. It’s the messenger. What’s really happening: Windows’ Timeout Detection and Recovery (TDR) watches your GPU, and if the card stops responding for a couple of seconds, Windows tries to reset it. When that recovery fails, you get the blue screen with nvlddmkm.sys attached. The underlying trigger is usually an unstable, corrupted, or conflicting driver, sometimes an overclock or power or heat problem, and occasionally genuine hardware fault.
Because crashes that happen outside gaming (idle at the desktop, during boot) can point past the driver, run the cheap diagnostics too. Windows Memory Diagnostic (type mdsched.exe and restart) checks for bad RAM. Monitor GPU temps with something like HWiNFO under load. Overheating past the low-to-mid 80s Celsius on a repeatedly stressed card is worth investigating.
When nothing has worked
You’ve done the DDU clean install, tried an older driver, cleared overclocks, checked memory and temps, and it still crashes. At that point, capture the evidence and escalate.
Your minidump files live in C:\Windows\Minidump. Note the exact NVIDIA driver version you’re running and the exact error code (VIDEO_TDR_FAILURE). With those two details, you can get a real answer instead of guesses. Post them to the NVIDIA subreddit or NVIDIA’s official driver feedback forum, or open a case with NVIDIA support directly. If the machine is under warranty and the crashes survive a stock, clean-driver configuration, that’s a strong sign of a hardware defect, and the seller or manufacturer should be your next call.
One more honest note: if your PC is brand new and blue-screened almost immediately, don’t assume Windows is corrupt. Suspect driver incompatibility, a BIOS or firmware update you’re missing, or a defective card, in that order.
If you remember one thing: the fix is a clean driver reinstall with DDU, not a plain update, because nvlddmkm.sys is the crash’s messenger, not its cause.