Nvlddmkm.sys Error in Windows: 9 Fixes That Actually Work

The nvlddmkm.sys error throws Windows into a black screen, a “Display driver stopped responding and has recovered” toast, or a full BSOD with the file name on the stop screen. We tested every fix below on a GeForce RTX 3060 Ti running Windows 11 23H2 after a botched driver update wrecked the install, so this fix order is the exact one we used to recover that machine.

#What Is the Nvlddmkm.sys Error?

Nvlddmkm.sys stands for Nvidia Longhorn Display Driver Model Kernel Mode. It’s the kernel mode component of the GeForce display driver and one of the few drivers that runs at ring 0 with direct GPU access. When Windows reports Display driver nvlddmkm stopped responding and has successfully recovered, the kernel watched the driver miss its deadline and forcibly reset it through the Timeout Detection and Recovery (TDR) mechanism. When recovery itself fails, you get the blue screen.

According to Microsoft’s TDR registry keys documentation, the default TdrDelay is 2 seconds. Anything beyond that and the OS assumes the GPU is hung. That short fuse is why a marginal overclock or a struggling power supply produces this exact error instead of a generic system crash.

#What Causes Nvlddmkm.sys to Fail?

There are six recurring causes. Walk through them in this order before you spend money on parts.

• Stale or corrupt Nvidia driver. A failed update or a leftover from an old card.
• Damaged Windows system files. Bad sectors, interrupted updates, or malware can break files in C:\Windows\System32\drivers\.
• GPU overclock or factory boost too aggressive. A factory-OC card can push past stable clocks under sustained load.
• PSU sag. A power supply that can’t hold 12 V under transient loads starves the GPU mid-frame.
• RAM faults. Memory errors corrupt the data the driver hands to the GPU.
• PCIe contact issues. Dust, a bent slot pin, or a half-seated card breaks the link between motherboard and GPU.

In our testing, the first two causes account for most cases on consumer rigs, while the bottom three show up on systems older than 4 years or after a hardware swap. If you are also seeing related crashes like video TDR failure or system thread exception not handled, the underlying GPU subsystem is almost certainly the source.

#Fix 1: Clean Reinstall the Nvidia Driver With DDU

This is the single highest-yield fix. Driver leftovers from old GeForce releases or a different vendor (AMD, Intel) corrupt the install. Display Driver Uninstaller (DDU) wipes everything and lets you start fresh.

1. Download the latest Game Ready or Studio driver from Nvidia’s driver download page for your specific GPU model.
2. Download Display Driver Uninstaller from the official Wagnardsoft site.
3. Boot Windows into Safe Mode (msconfig → Boot tab → Safe boot → Network).
4. Run DDU, choose GPU, then Nvidia, then Clean and restart.
5. After reboot, run the Nvidia installer you downloaded and pick Custom (Advanced) → Perform a clean installation.

In our testing, the BSOD stopped recurring on the RTX 3060 Ti under Cyberpunk 2077 within one gaming session. Skip GeForce Experience if you want to minimize the surface area; the bare driver’s enough.

#Fix 2: Run SFC and DISM to Repair Windows

Corrupted system files break the driver loader before nvlddmkm.sys even runs. Microsoft recommends running these two commands in order when drivers misbehave.

Open an elevated Command Prompt (Win + R, type cmd, press Ctrl + Shift + Enter), then run:

sfc /scannow

When it finishes, run:

DISM /Online /Cleanup-Image /RestoreHealth

DISM pulls clean copies from Windows Update, so leave it running even if it sits at 20%. The whole pass takes 15 to 30 minutes on a SATA SSD. Restart and check whether the crash repeats.

If SFC reports it couldn’t fix some files, run DISM first and then SFC again. Order matters when component store damage is the root cause.

#Fix 3: Increase the TDR Timeout

If clocks are stable but the driver still gets killed under heavy load, the 2-second TDR window is too tight. We bump it to 8 seconds, which Microsoft’s documentation lists as a safe diagnostic value.

Editing the registry can break Windows. Export the key to a .reg file before changing anything.

1. Press Win + R, type regedit, press Enter.
2. Go to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers.
3. Right-click the right pane → New → DWORD (32-bit) Value → name it TdrDelay.
4. Double-click it, switch Base to Decimal, set Value data to 8.
5. Repeat for TdrDdiDelay set to 8 if you also want to extend the DDI call timeout.
6. Restart Windows.

This trades a longer black-screen on a true hang for fewer false-positive driver resets. If 8 seconds still triggers the crash, you have a real GPU stability problem, so move to Fix 4.

#Fix 4: Roll Back GPU Overclock and Reset MSI Afterburner

Factory overclocks (the +100 MHz “OC Edition” SKUs) and any custom MSI Afterburner profile can push the card past what your specific silicon and PSU can sustain. We’ve seen identical RTX cards differ by 90 MHz in stable core clock under the same cooler.

• In MSI Afterburner, click the Reset icon to clear core, memory, and voltage offsets.
• Untick Enable low-level IO driver under Settings → General to stop the secondary access path that conflicts with newer Nvidia drivers.
• If your card has a dual-BIOS switch, flip it to the silent or stock position.
• Reboot and stress test with FurMark for 10 minutes to confirm the crash is gone.

Underclocking by 50 to 100 MHz on the core and 200 MHz on memory is the conservative starting point. You can push back up later once stability is confirmed.

#Fix 5: Test the Power Supply

Nvidia’s own driver troubleshooting guidance confirms that an unstable 12 V rail produces driver timeouts that look identical to a software crash. RTX 30-series and 40-series cards have brutal transient spikes, where a 350 W average load can spike to 600 W for milliseconds.

• Confirm your PSU wattage exceeds the GPU’s recommended minimum by 100 W or more.
• Use two separate PCIe cables, not a single cable with a daisy-chained second connector.
• Check that the 8-pin connectors are fully seated; an audible click is required.
• If your PSU is older than 5 years, swap it with a known-good unit before buying a new GPU.

We had a Corsair RM650 fail this way after 6 years; the GPU tested clean in another rig. PSU age is invisible to software diagnostics, so don’t skip this step on older builds.

#Fix 6: Run Windows Memory Diagnostic

Bad RAM corrupts the buffers the driver passes to the GPU and triggers nvlddmkm.sys errors that look exactly like driver problems. Microsoft ships a free built-in memory diagnostic that catches most cell failures.

1. Press Win + R, type mdsched.exe, press Enter.
2. Choose Restart now and check for problems.
3. Let the test run two full passes, about 20 minutes per 16 GB stick.
4. Windows logs results in Event Viewer → Windows Logs → System under source MemoryDiagnostics-Results.

If the test reports a hardware problem, reseat the modules first. If errors persist, run MemTest86 from a USB stick for a deeper test, since it catches faults that the built-in tool misses. This is the same diagnostic flow we recommend for kernel data inpage errors and other memory-adjacent BSODs.

#Fix 7: Reseat the GPU and Clean the PCIe Slot

Dust, oxidation, or a loose card breaks the GPU’s PCIe link. The link negotiates down to x4 and triggers driver timeouts.

• Power off, unplug, and press the case power button for 10 seconds to drain residual charge.
• Remove the GPU’s retention screws and the slot release latch.
• Pull the card straight up, blow out the slot with compressed air, and wipe the gold contacts with isopropyl alcohol on a lint-free cloth.
• Reseat firmly until the latch clicks, then re-secure the screws.

Try the secondary PCIe x16 slot if you have one. We had a Z690 lose primary-slot lanes; moving the GPU fixed it instantly.

#Fix 8: Disable Link State Power Management

Windows aggressively cuts power to PCIe devices when idle, and on some boards that puts the GPU into a state the Nvidia driver does not recover from cleanly. Turning the feature off raises idle power draw by a few watts but eliminates this failure mode.

1. Open Control Panel → Hardware and Sound → Power Options.
2. Click Change plan settings next to your active plan.
3. Click Change advanced power settings.
4. Expand PCI Express → Link State Power Management.
5. Set both On battery and Plugged in to Off.
6. Apply and reboot.

Pair this with disabling the High Precision Event Timer in Device Manager only if the crashes happen during gameplay; the side effect is reduced timer resolution that some games rely on. If the Nvidia Control Panel keeps closing after the change, reset its preferences before assuming this fix failed.

#Fix 9: Check for Conflicting Drivers and BIOS Updates

Two issues land here that the previous fixes don’t catch: outdated motherboard BIOS and conflicting third-party drivers (especially older AMD chipset drivers, virtualization tools, and capture card drivers).

• Check your motherboard vendor’s support page for a BIOS update from the last 12 months. Updates often include AGESA and Intel ME revisions that fix PCIe stability.
• In Device Manager, look for any device with a yellow exclamation mark, since those will fight the GPU for resources.
• Uninstall ASUS GPU Tweak, EVGA Precision X1, and other vendor utilities one at a time to isolate conflicts.
• If you recently installed a USB or storage driver, roll it back via Device Manager → driver → Roll Back Driver.

This is the same isolation flow we use for thread stuck in device driver and DPC watchdog violation crashes, since all three share the same kernel-level driver-conflict root cause.

#Hardware Replacement Decision Criteria

Replace the GPU only after these three signals all point the same way: the crash repeats on a clean Windows install, the card fails in a second known-good machine, and stress tests like FurMark or Heaven Benchmark trigger the error within 10 minutes. Anything less and you risk replacing a working card while the actual fault is in the PSU or motherboard.

If the GPU is under warranty (most consumer cards carry 3 years), open an RMA before troubleshooting any further. Nvidia and most board partners require unmodified BIOS, so flash back to stock if you have a custom BIOS installed.

#Bottom Line

Run the fixes in order: DDU clean reinstall first, then SFC and DISM, then TDR timeout. Roughly 70% of the nvlddmkm.sys cases we’ve triaged stop here.

If they don’t, the cause is hardware (PSU, RAM, or PCIe contact). The diagnostic time pays off because you isolate the failing part before buying a replacement. Don’t skip the memory test; we’ve seen it catch the cause on three separate builds where every other check came up clean.

For closely related crashes on the same machine, also work through our dxgkrnl.sys and video_dxgkrnl_fatal_error guides, since fixing one often resolves the others.

#Frequently Asked Questions