nvlddmkm Error (Event ID 153): Why Your NVIDIA Driver Crashes Games, and the Fixes That Work
If your game freezes, your screen goes black for a few seconds, and then the desktop comes back, Event Viewer may show Event ID 153 from source nvlddmkm. That entry alone does not tell you which part failed. It tells you Windows reset your NVIDIA driver after a timeout. This guide shows you how to read that entry, match your crash pattern to the right test, and work through the fixes in an order that does not waste your time.
Work one change at a time. After each change, test with the same game or app that triggered the crash, then check Event Viewer again for new Event ID 153 entries. If you change five things at once, you will never know which one helped.
In This Guide
- What Event ID 153 Means
- Read the Log Before You Change Anything
- Match Your Pattern to the Next Test
- Quick Checks That Cost Nothing
- Test for a Driver Regression
- Clean Driver Install Without DDU First
- DDU as a Later Step, Not the First Step
- Stock Clock Test for Overclocks and Memory Profiles
- Power, Heat, and Connection Checks
- TDR Delay and Why Disabling TDR Is a Bad Idea
- When to Suspect Hardware Instead of Software
- FAQ
- Bottom Line
What Event ID 153 Means
nvlddmkm is the NVIDIA display driver file. When Windows sees the graphics card fail to respond in time, it resets that driver so the desktop can recover. That reset is called Timeout Detection and Recovery, or TDR. The default timeout is 2 seconds. If the reset works, you see a flicker or a black screen for a few seconds, then the desktop returns. You may also see the message “Display driver stopped responding and has recovered”. If the reset fails, the result can be a blue screen with VIDEO_TDR_FAILURE.
The full Event Viewer line often reads “The description for Event ID 153 from source nvlddmkm cannot be found.” That sentence is not a second error. It means Windows has no stored description text for this event. Do not search for a fix for the missing description. The useful part of the entry is the driver line underneath it, which names the device and the type of reset, for example “Error occurred on GPUID” or “Reset TDR occurred on GPUID”.
Two details in that line do not help you diagnose the fault. The Video device number, such as Video3, and the GPUID number differ between PCs. They are not fault codes, and they do not point to one failed part. What matters is the pattern around the entry: when it happens, which app was running, which driver version was installed, and whether it repeats.
Event ID 153 itself is not documented by Microsoft or NVIDIA as a standalone error, based on checks reported on October 15, 2026. The TDR mechanism behind it is documented by Microsoft. Treat Event ID 153 as a clue that a timeout and reset happened, not as a diagnosis that your card is dead or that one driver version is at fault.
Read the Log Before You Change Anything
Start in Event Viewer. This takes about ten minutes, and it stops you from reinstalling drivers for a problem that only happens in one game.
- Right-click Start and choose Event Viewer.
- Open Windows Logs, then System.
- Click Filter Current Log on the right.
- In Event sources, select or type nvlddmkm.
- Look for Event ID 153 entries and note the exact time of each one.
Open each entry and read the driver line in the middle of the text. Ignore the first and last paragraphs, which repeat on every entry. Write down four things for the most recent crash: the time, the app that was running, your current driver version, and any entries logged at the same second, such as nvlddmkm Event ID 13 or 14, a Display source Event ID 4101, or a LiveKernelEvent 141 or 117.
For a timeline view, open Reliability Monitor. Press Windows plus R, type perfmon /rel, and press Enter. Reliability Monitor shows crashes and driver events on a calendar, which makes it easier to see whether the problem started on the day of a driver update, a Windows update, or a game patch.
Windows may also store live kernel dumps in C:\Windows\LiveKernelReports. You do not need to open those files to follow this guide. Their presence at the same time as an Event ID 153 entry is extra evidence that a GPU timeout and recovery happened. Do not send dump files to anyone except a vendor support channel that has asked for them for a specific case.
Match Your Pattern to the Next Test
Use this table to pick your next test. A pattern points to a test. It does not prove a cause.
| Pattern you see | Test next |
|---|---|
| Crashes started right after a driver update | Test for a driver regression. See the driver regression section below. |
| Crashes happen in one game only | Check that game first. Verify its files, check its patch notes, and test with overlays off before you touch the driver. |
| Crashes happen in every game, and in video playback | Work the driver steps in order: clean install, then DDU if needed, then the stock clock test. |
| Crashes happen at idle, in a browser, or when alt-tabbing | Check the driver regression section and the power management test. Power state changes are a reported trigger, not a proven cause for every PC. |
| Crashes come with artifacts, strange colors, or a full shutdown | Stop at the power, heat, and connection checks. Do not keep reinstalling drivers for a pattern that looks like hardware. |
Quick Checks That Cost Nothing
Do these checks before any driver reinstall. Change one thing, test, and check the log again.
Test with the same trigger. Use the same game, the same settings, and roughly the same workload that crashed before. A test that never loads the GPU proves nothing.
Close overlays as a test. Turn off in-game overlays one at a time, including the NVIDIA overlay and any third-party overlay you run. If the crashes stop, re-enable overlays one at a time to find the one that matters. An overlay is a suspect, not a proven cause, until this test changes the result.
Clear the shader caches. Stale shader cache files can cause crashes after a driver update or a major game patch. Open Settings, then System, then Storage, then Temporary files. Check DirectX Shader Cache and click Remove files. Then press Windows plus R, type %localappdata%, and press Enter. Open the NVIDIA folder if present and delete the contents of folders such as DXCache and GLCache. If Windows says a file is in use, skip it. If a folder does not exist, skip it. NVIDIA has changed cache locations across driver versions, and Windows recreates these files as needed.
Check what else changed that day. Look at Windows Update history and the game patch history for the same date as your first Event ID 153 entry. A driver update that arrived through Windows Update can be the same regression as a manual driver update. Note it, and do not assume the newest driver is always the safest choice.
Set power management as a test, not as a permanent fix. In NVIDIA Control Panel, under Manage 3D Settings, you can set Power Management Mode to Prefer Maximum Performance for one game profile. This keeps the card from dropping to a low power state as aggressively. Test with it. If it changes nothing, set it back. On a laptop, this setting costs battery life and adds heat, so do not leave it on without a reason.
Test for a Driver Regression
A driver regression means a new driver version introduced a problem that the previous version did not have. This section is date-stamped, because driver versions move fast. Check the NVIDIA driver page again before you act on it.
As checked on October 28, 2026, GeForce Game Ready Driver 621.01 was released on October 21, 2026. VideoCardz reported on October 24, 2026 that NVIDIA had acknowledged reports of video playback freezes and black screens on 621.01, had filed a bug report, and was collecting diagnostic data. Those reports describe freezes in browsers and media players, and in some cases when a game starts while a video plays on another monitor. Most reports named RTX 5080 cards, with at least one RTX 4090 report. No fix date was announced in those reports, and the exact cause was not confirmed.
Multiple users reported that rolling back to GeForce Game Ready Driver 617.42, released on October 6, 2026, stopped the freezes. Treat that as a user-reported workaround, not as a confirmed cure. It applies to the 621.01 video freeze pattern. It does not prove that every Event ID 153 crash is a 621.01 regression, and a small number of users reported similar symptoms on older drivers.
Use this test only if all three points fit your case: you are on 621.01, your crashes match the video freeze or black screen pattern, and your crashes started after the 621.01 update. If those points fit, roll back with the NVIDIA installer or with Device Manager, test with the same trigger, and check for new Event ID 153 entries. If a newer driver than 621.01 is available when you read this, check its release notes first. A newer driver may already list this fix, and rolling back to an older driver would then be the wrong move.
If you roll back, stop the NVIDIA app and Windows Update from offering 621.01 again right away. Rolling back and then updating an hour later puts you back where you started. Wait for a driver release that lists the fix before you update again.
Clean Driver Install Without DDU First
A clean install removes old driver settings without a third-party tool. Try this before DDU.
- Download the driver for your exact card from the NVIDIA driver download page. Save the installer where you can find it after a restart.
- Close all games and GPU tools.
- Run the installer and choose Custom (Advanced), not Express.
- Check Perform a clean installation.
- Finish the install and restart the PC.
- Test with the same game or app that crashed before, for at least 15 to 30 minutes, then check Event Viewer for new Event ID 153 entries.
If you use a laptop, prefer the driver package from your laptop maker support page for your exact model, where one exists. Laptop makers ship drivers tuned for that model power and thermal profile. A generic desktop driver package can refuse to install on a laptop, or it can install and behave differently. Use the generic NVIDIA package only if your laptop maker has no suitable package.
If the crashes stop after the clean install, reapply only the NVIDIA settings you actually need, one at a time. If you restore a full profile in one step, you may restore the setting that caused the problem.
DDU as a Later Step, Not the First Step
Display Driver Uninstaller, or DDU, removes driver files, services, and leftover installer artifacts more thoroughly than the NVIDIA installer clean option. It is a third-party community tool, not an NVIDIA tool. Use it only when a clean install did not fix the problem, or when installs keep failing. Do not start with DDU for every driver problem.
Our full steps are in our DDU clean install guide. The short version is: download the NVIDIA driver first, boot Windows into Safe Mode, run DDU, select NVIDIA, choose Clean and restart, then install the driver you downloaded. Test with the same trigger app, then check Event Viewer again.
Do not run DDU and then leave the PC without a driver while you search for one. Download first, clean second, install third. That order avoids a long stretch with a basic display driver and avoids Windows Update installing a driver version you did not choose.
Stock Clock Test for Overclocks and Memory Profiles
An overclock that passed last month can fail after a driver update, a game patch, or a hotter room. The stock clock test costs nothing and it separates a stability problem from a driver problem.
Reset the GPU to stock clocks. Open your GPU tuning tool and reset it to defaults. If your card shipped with a factory overclock, use the stock or reference clock option in your tuning tool if one is offered. Do not set a new overclock while you test. Test with the same trigger app, then check the log.
Turn off XMP or EXPO as a test. XMP on Intel boards and EXPO on AMD boards are memory profiles that run RAM above its base speed. That is a form of overclocking, even when the profile came with the RAM kit. Restart into BIOS, turn the profile off, save, and test again. If the crashes stop, you have found a stability problem, not proof that the RAM is faulty. RAM can be fine at base speed and unstable at a profile speed on a given board and BIOS version. Re-enable settings one at a time after the test, and check your motherboard maker guidance before you settle on a permanent setting.
Read the result correctly. If stock clocks stop the crashes, the problem was stability at the old settings, not proof that the card or the RAM is broken. If stock clocks change nothing, move on to the power, heat, and connection checks. Do not keep lowering clocks in search of a fix. That hides the real cause instead of finding it.
Power, Heat, and Connection Checks
Software fixes cannot fix a loose cable, a weak power supply, or a card that overheats under load. Check these items with the PC off where the step involves touching hardware.
Check power cables. With the PC off and unplugged, check that the power cables to the card are fully seated. On cards with a 16-pin connector, check that the connector is fully inserted and not strained at an angle. A loose connector can cause voltage drops that look like driver crashes. Do not force a connector, and do not use an adapter you do not trust.
Check PSU capacity against the card maker recommendation. Find the recommended PSU wattage on your card maker product page for your exact model. Compare it with your installed PSU wattage. A PSU that is far below the recommendation is a suspect, not a proven fault. A PSU that meets the recommendation is not automatically cleared either, because age and quality matter, but wattage is the check you can do without buying parts.
Check temperatures against your model, not against one number. Safe GPU temperatures are model-specific. A temperature that is normal for one card can be high for another. Check your card maker specification page, then monitor temperatures during the same workload that crashed. Our GPU temperature guide covers how to monitor and what to compare against. Do not use one threshold from another guide as a verdict for your card.
Check a riser cable only if you use one. Riser cables can degrade the PCIe signal, and that signal problem can trigger driver resets. If you use a riser, test the card directly in the motherboard slot where you can do so safely. If the crashes stop without the riser, the riser is the suspect. If you cannot test safely, treat the riser as unproven, not as the cause.
Check the monitor path last in this section. Try a different cable and a different port as a test, especially if crashes happen when a second monitor is connected or when a video plays on a second screen. A cable fault is uncommon for Event ID 153, but it is cheap to rule out, and it avoids a driver reinstall for a display path problem.
TDR Delay and Why Disabling TDR Is a Bad Idea
You will find guides that tell you to edit the Windows registry to raise the TDR delay. Understand what that edit does before you try it.
The relevant keys sit under HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\GraphicsDrivers. The defaults are TdrDelay at 2 seconds, TdrDdiDelay at 5 seconds, and TdrLevel at 3, which means recover on timeout. TdrDelay is how long the GPU may delay a preemption request before Windows treats it as frozen. Raising it gives the GPU more time. It does not fix a bad driver, an unstable overclock, or failing hardware. It only changes when Windows gives up and resets.
Do not set TdrLevel to 0. That disables TDR detection. A hung GPU that cannot be reset becomes a hard lock that needs a power cycle, and you lose the recovery that was saving your desktop. If you try a TdrDelay change at all, back up the registry first, record the old value, change it in a small step, and test. If it changes nothing, set it back. This is a last test, not a fix, and it belongs after the driver, stock clock, power, and heat checks, not before them.
When to Suspect Hardware Instead of Software
Stop changing software when the software tests are exhausted. Suspect hardware, and get help, when most of these points are true:
- Crashes survive a clean driver install, DDU, and a known good driver version.
- Crashes survive the stock clock test, with XMP or EXPO off.
- Crashes happen across different games and apps, not in one title.
- Crashes come with visual artifacts, strange colors, or full shutdowns.
- Crashes happen under load and also at idle, with no pattern tied to one update.
None of those points alone proves a hardware fault. Together, they mean further driver reinstalls are unlikely to help. Do not swap parts on guesswork. Check whether the card or the PC is still under warranty, keep your Event Viewer notes and driver version history, and get a professional check if you cannot test a part safely. A shop can test a card in a known good PC, which is the test a driver guide cannot give you.
If your crashes look like stuttering rather than a driver reset, read our game stuttering guide next. Stuttering and a TDR reset are different problems, and they need different fix lists.
FAQ
What does “cannot be found” mean in Event ID 153?
It means Windows has no stored description text for the event. It is not a second error, and there is no description pack to install. Read the driver line underneath it instead.
Does Event ID 153 mean my graphics card is dead?
No. It means Windows reset the driver after a timeout. A dead card is only one possible cause, and it is not the most common one. Work the ordered tests before you suspect hardware.
Should I roll back to driver 617.42?
Only if you are on 621.01, your crashes match the video freeze or black screen pattern, and they started after the 621.01 update. Even then, treat rollback as a test. Check for a newer driver first, because a newer release may already list the fix.
Should I use DDU first?
No. Try a clean install with the NVIDIA installer first. Use DDU when a clean install did not work, or when installs keep failing. DDU is a third-party tool, not an NVIDIA tool.
Should I raise the TDR delay?
Only as a last test, with a registry backup, and never by disabling TDR. Raising the delay gives the GPU more time. It does not fix a bad driver or unstable hardware.
Why does Event ID 153 happen at idle or in a browser?
Power state changes are a reported trigger, because the card switches power states quickly when you browse, alt-tab, or start a video. That is a reported pattern, not a proven cause for every PC. Check the driver regression section first if the idle crashes started after a driver update.
Bottom Line
Event ID 153 means Windows reset your NVIDIA driver after a timeout. It does not name the failed part. Read the log, match your crash pattern to the next test, and change one thing at a time. Test for a driver regression first if the crashes started with a driver update. Try a clean install before DDU, run the stock clock test before you blame hardware, and never disable TDR to hide the problem. When clean drivers, stock clocks, and a known good driver version all fail, and crashes appear across games with artifacts or shutdowns, stop reinstalling drivers and suspect hardware instead.
