How to Fix VIDEO_TDR_FAILURE (nvlddmkm.sys)

VIDEO_TDR_FAILURE is the blue screen that loves to strike at the worst moment — in the middle of a game, a render, or a video call — and it usually names a driver file like nvlddmkm.sys, atikmpag.sys, or igdkmd64.sys. In the fifteen-plus years I've run Dave's Computers in Somerville, NJ, this is one of the most common crashes gamers and video editors bring through the door, and I've seen more of it lately — partly because Windows has been quietly swapping and downgrading GPU drivers during updates, which is a classic way to trigger it.

The good news: most VIDEO_TDR_FAILURE crashes are a driver problem, and a proper clean reinstall fixes them. The catch is that "proper" matters — the lazy reinstall most people do is exactly why it keeps coming back. And a stubborn minority of these are hardware: an overheating or failing graphics card, or a power supply that can't keep up under load. This guide covers what the error means, how to read which GPU is involved, some steps you can try in the order I'd work through them, and the point where guessing gets expensive and it's smarter to have it tested.

What VIDEO_TDR_FAILURE actually means

VIDEO_TDR_FAILURE shows the stop code 0x00000116. TDR stands for Timeout Detection and Recovery — a built-in Windows feature that watches your graphics card. If the GPU stops responding for a couple of seconds, Windows tries to reset the display driver on the fly to avoid a crash. You've probably seen this work: the screen flashes black for a moment and recovers with a little "display driver stopped responding and has recovered" message.

VIDEO_TDR_FAILURE is what you get when that rescue fails — Windows tried to reset the graphics driver, couldn't get the GPU talking again, and shut the system down to protect it. So the real question is never "what is TDR," it's "why did my graphics card stop responding in the first place?" The driver file named on the screen is your first clue, because it tells you which GPU is involved.

Which driver file means which graphics card

The file listed under the stop code points at the graphics driver that fell over:

Driver file	What it points to
nvlddmkm.sys	NVIDIA GeForce / RTX driver — by far the most common one I see named.
atikmpag.sys / atikmdag.sys	AMD Radeon driver.
igdkmd64.sys / igdkmd32.sys	Intel integrated graphics driver (common on laptops).
dxgkrnl.sys / no file named	The generic Windows graphics stack — usually still points back to your GPU driver or the card.

If your laptop has both Intel integrated graphics and a separate NVIDIA or AMD chip, you can see either file named depending on which one was driving the screen when it crashed — often when you've got a second monitor plugged in over HDMI or USB-C.

The causes we actually see on the bench

Here's how these break down in real life, roughly most common to least:

1. A corrupted, mismatched, or outdated GPU driver. The number-one cause. It spikes after a messy driver update installed on top of an old one, or after a Windows update silently rolls your driver back to a version that doesn't match your card or your games.
2. Overheating. A graphics card running too hot — clogged fans, dried-out thermal paste or pads, poor case airflow — will time out under load. A prime suspect when crashes cluster during gaming or rendering but never at idle.
3. An unstable overclock or factory boost. Too-aggressive core/memory clocks, or an undervolt that isn't actually stable, will throw 0x116 the moment the card is pushed. This includes some factory-overclocked cards that aren't stable out of the box.
4. Power delivery problems. A failing or underpowered PSU sags under load and starves the GPU; a loose or daisy-chained PCIe power cable does the same. Very common on gaming rigs that crash only during demanding scenes.
5. A failing graphics card or bad VRAM. A card on its way out — artifacting, bad memory, age — surfaces as repeated TDR crashes that survive every driver fix. The one nobody wants to hear.
6. A genuinely buggy driver release. It happens. Sometimes the fix is rolling back to the last stable version and waiting for the next one.
7. A loose card or riser. A GPU not fully seated in its slot, or a flaky riser cable in a vertical-mount build, can cause it too.

How to fix VIDEO_TDR_FAILURE — some steps you can try

Here are some steps you can try yourself, in the order I'd work through them. Most people get fixed somewhere in the first few — and if you reach the end and it's still crashing, that's the section just below, where it stops being a safe do-it-yourself job. Retest after each step; the moment the crashes stop, you've found it.

1. Confirm which GPU crashed

Note the driver file on the blue screen, then open Event Viewer → Windows Logs → System and look for Display or nvlddmkm errors around the crash time. A minidump opened in BlueScreenView confirms the culprit. This tells you whether you're chasing the NVIDIA/AMD card or the Intel integrated graphics.

2. Do a clean driver reinstall with DDU — the step most people skip

This is the single highest-value fix, and doing it properly is what separates a real fix from a crash that comes back in a week. Installing a new driver on top of a corrupted one usually leaves the corruption in place. Instead:

1. Download the latest driver from NVIDIA / AMD / Intel (don't install yet)
2. Download Display Driver Uninstaller (DDU)
3. Boot into Safe Mode
4. Run DDU and choose "Clean and restart" for your GPU brand
5. Back in normal Windows, install the fresh driver you downloaded

Here's how to get into Safe Mode if you're not sure. A clean DDU reinstall resolves the large majority of these.

3. Roll back the driver if it started right after an update

If the crashes began immediately after a driver or Windows update, go to Device Manager → Display adapters → your GPU → Driver → Roll Back Driver, or install the previous stable version manually. Given how often Windows Update re-installs an older driver behind your back, this is worth checking early.

4. Check temperatures under load

Install HWiNFO or GPU-Z and watch your GPU temperature while gaming or running a render. If it's running hot, clear out dust, confirm the fans actually spin, and consider fresh thermal paste or pads. Heat is one of the most common — and most fixable — triggers. (Our notes on how cooling affects performance apply to desktops too.)

5. Turn off overclocking

If you (or the card's factory profile) are overclocking, set the GPU back to stock in MSI Afterburner or your tuning tool and test. If a factory-OC card is unstable, a small underclock on the memory often settles it.

6. Reseat the card and check power

On a desktop, reseat the GPU in its slot and make sure each PCIe power connector is fully plugged in — ideally using separate cables rather than one daisy-chained run. If the blue screens cluster under heavy load, an aging or undersized power supply is a strong suspect.

7. Test the card itself

Once drivers, heat, and clocks are ruled out, a controlled stress test (FurMark or 3DMark) forces the fault to show, and swapping in a known-good GPU confirms whether the card is the problem. That's bench work we do every week, especially on custom and gaming PCs.

How much does it cost to fix VIDEO_TDR_FAILURE in NJ?

I can't quote this one blind, because 0x116 can be a fifteen-minute clean driver reinstall or a failing graphics card — very different jobs. What I can tell you is how our pricing works, with no surprises.

Plenty of these are a quick, inexpensive driver fix; a repaste, PSU, or GPU swap is more involved. Either way, nothing gets ordered or replaced without your okay. Here's how our diagnostic works, and a broader look at what computer repair costs in New Jersey.

Frequently asked questions