How to fix "The description for Event ID 0 from source nvlddmkm cannot be found"?

"The description for Event ID 0 cannot be found" means Event Viewer has no message text for that event number from that source. It is not proof that anything is missing or corrupted. When the source is nvlddmkm, the NVIDIA display driver, the useful part is the line under the boilerplate, and it usually sits next to a graphics driver timeout that Windows had to recover from.

Facts checked on 14 October 2026. Steps written for Windows 11 (25H2 and 26H2) and Windows 10 22H2 with an NVIDIA GeForce graphics card.

What "the description for Event ID 0 cannot be found" means

A program that writes to the Windows event log does not store a full sentence. It stores its name (the source), an event number and a few raw values. Event Viewer builds the readable text later: it looks up the source in the registry, opens the message file registered there and takes the sentence that belongs to that number.

Microsoft documents this under event sources. Each source has a key under HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\EventLog, inside the log it writes to, and the value EventMessageFile holds the path to its message file. When the source has no such file, Microsoft's own words are that Event Viewer "cannot map any event identifiers or event categories to a description string, and will display an error". That error is the text you are reading.

The full entry as users post it (here with ID 153; ID 0, 13 and 14 look the same)

The description for Event ID 153 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer.

If the event originated on another computer, the display information had to be saved with the event.

The following information was included with the event:

\Device\Video3
Error occurred on GPUID: 100

The message resource is present but the message was not found in the message table

• Event ID 0 is only a number. Microsoft says each event source defines its own numbered events. A program that never numbers its events writes all of them with 0. The number carries no meaning across programs: 0 from one source and 0 from another are unrelated.
• "Not installed or corrupted" is a guess by Windows. The sentence is the same for every source and every number. It lists the two reasons Windows can think of. It has not checked the NVIDIA driver.
• The last line settles it. "The message resource is present but the message was not found in the message table" is Windows error 15027 (ERROR_EVT_MESSAGE_NOT_FOUND). It says the message file was found and opened, and this number is simply not in it.
• The middle is the real content. Everything after "The following information was included with the event" are the raw values the driver stored. Event Viewer shows them unformatted because it has no sentence to put them in.

What nvlddmkm is and what its events 0, 13, 14 and 153 tell you

nvlddmkm is the kernel part of the NVIDIA graphics driver, the file nvlddmkm.sys. Microsoft's debugging pages show it as the module named in the stop error VIDEO_TDR_FAILURE on PCs with an NVIDIA card. In Event Viewer it appears as a source in the System log, because device drivers log there.

NVIDIA publishes no list of these Windows event numbers that we could find, so nobody outside NVIDIA can decode an ID with certainty. What can be read is the extra line. The table shows the lines users post most often on Microsoft Q&A, together with what was happening on their PCs. Treat the right column as reports, not as a vendor definition.

• Count incidents, not entries. A single crash often leaves five to fifteen nvlddmkm entries with the same time stamp.
• The number after \Device\Video differs between PCs. We could not confirm from Microsoft or NVIDIA what it encodes, so do not read a fault into Video3 or Video5.
• The same line every time suggests one repeating failure. Lines that change from crash to crash are more typical of unstable hardware.
• No symptom, no problem. If the entries do not line up with any flicker, black screen or crash, there is nothing to repair. The missing description alone is harmless.

What Windows does when the graphics card stops answering (TDR)

The crashes that come with these events are almost always a timeout. Windows watches the graphics card. If the card takes too long with one task, Windows declares it frozen, resets the driver and the card, and tries to give you the desktop back. Microsoft calls this Timeout Detection and Recovery, TDR.

This explains the names you meet next to the nvlddmkm events. LiveKernelEvent 141 in Reliability Monitor is the live dump 0x141: one engine of the card did not answer in time and was reset. LiveKernelEvent 117 is the timeout of the whole adapter. VIDEO_TDR_FAILURE means the reset itself failed or the timeouts came too fast.

The causes behind nvlddmkm errors, in the order to rule them out

Nobody publishes how often each cause occurs, so we do not invent percentages. The order below is the order in which NVIDIA's support asks you to rule things out before it accepts a crash report, which is also the order from cheapest to most expensive.

How to fix the crashes behind Event ID 0 from source nvlddmkm

Start with Fix 1 in every case: it takes ten minutes and tells you which of the others you need. The fixes are ordered from free and reversible to those that open the PC case. Stop as soon as the crashes stop.

Fix 1. Read the nvlddmkm events and the crash entries next to them

10 minutesNo riskEveryone. Do this before you change anything.

The event text is useless, but three facts are not: the extra line of each event, how many separate incidents there were, and whether Windows recorded a recovered timeout or a blue screen at the same minute. They decide which fix to use.

1. Right-click Start and select Terminal (Admin). On Windows 10, open PowerShell as administrator instead.
2. Paste this command and press Enter. It lists the last 50 events of the NVIDIA driver with the raw values the driver stored.

   Get-WinEvent -FilterHashtable @{ LogName='System'; ProviderName='nvlddmkm' } -MaxEvents 50 |
     Select-Object TimeCreated, Id, LevelDisplayName, @{ n='Data'; e={ $_.Properties.Value -join ' ' } } |
     Format-List

3. Group the entries by time. Entries within the same few seconds are one incident. Write down the date, the time, the Id and the Data line of each incident.
4. Press Win + R, type perfmon /rel and press Enter to open Reliability Monitor.
5. Select the days of your incidents and open the technical details of each critical entry. A recovered timeout carries the problem event name LiveKernelEvent with the code 141 or 117. We could not confirm the exact wording of the entry title from Microsoft.
6. Open File Explorer and look into %SystemRoot%\LiveKernelReports and %SystemRoot%\Minidump. Dump files dated at the time of an incident confirm a timeout or a blue screen.

Fix 2. Put the graphics card, the processor and the memory back to default speeds

10 to 20 minutesLow riskAny part is overclocked, a tuning tool is installed, or an XMP or EXPO memory profile is on. Also worth ten minutes when you are not sure.

NVIDIA asks for exactly this before it looks at a crash: its support article says some programs are more sensitive to overclocking than others, so a setting that survives one game can time out in another. Microsoft lists over-clocked components first among the hardware reasons for a GPU timeout.

1. Close every tuning tool for the graphics card, set its clock, voltage and fan curves back to default, and stop it from starting with Windows: Stop apps from opening at startup.
2. Open the NVIDIA app, go to System and then Advanced, and turn on Debug Mode. NVIDIA's article says the same switch is in the menu bar of the NVIDIA Control Panel, without naming the menu.
3. Restart the PC and enter the firmware setup (BIOS or UEFI). The key is shown at start-up and differs between makers.
4. Set the memory profile from XMP (Intel) or EXPO (AMD) to Default, as NVIDIA's article describes, and remove any manual processor overclock.
5. Save, start Windows and play the game that crashed, for at least as long as it used to take to fail.

• Stable now: one of the settings was the cause. Turn them back on one at a time, memory profile first, and test after each.
• Stable only with Debug Mode on: the card cannot hold its factory overclock any more. That is a warranty matter for the card maker.
• Still crashing: the clock speeds are ruled out. Leave everything at default while you continue, so the next tests are clean.

Fix 3. Install the current NVIDIA driver with a clean installation

15 minutesLow riskThe crashes did not begin with a specific driver update, or the driver was installed while games and other programs were running.

NVIDIA's support article gives the reason: when a driver is installed while other programs run in the background, some processes may not be released properly, which leads to installation problems. A clean installation replaces the files and resets the driver's settings to defaults.

1. Open NVIDIA's driver download page and pick your graphics card and your Windows version. On the day of checking the newest Game Ready driver was 617.42, released on 6 October 2026.
2. Download the installer. Do not start it yet.
3. Close games, launchers, browsers, recording and monitoring tools. Check the notification area for programs that still run.
4. Start the installer. When it asks for the kind of installation, select Custom (Advanced).
5. Tick Perform a clean installation and continue. These are the two labels NVIDIA's support article names.
6. Restart the PC when the installer finishes, even if it does not ask.

• Which cards the current driver supports. The 617.42 release notes list desktop GeForce cards from the GTX 16 series to the RTX 50 series. If your card is older, the download page offers the last driver made for it.
• Laptops. NVIDIA's release notes say notebook issues can be system-specific. If the NVIDIA driver keeps failing on a laptop, try the graphics driver from the laptop maker's support page.
• A deeper clean. NVIDIA's article also names the third-party tool Display Driver Uninstaller as a way to remove all NVIDIA software before installing. NVIDIA calls it a tool for advanced users who know Safe Mode. It is not an NVIDIA product.

Fix 4. Go back to the driver that worked

10 minutesLow riskThe first incident in your list from Fix 1 is on or just after the date of a driver update.

A single driver version can break one game or one family of cards, and NVIDIA's release notes list such faults as fixed in later versions. Going back one version tests this directly: if the crashes stop, the driver was the cause and the hardware is fine.

1. Open Device Manager. Right-click Start and select Device Manager.

   

2. Open the device's Driver tab. Expand the category (for example Display adapters), right-click the device, select Properties and open the Driver tab. Driver Date and Driver Version show what is installed now.

   

3. Update the driver. Select Update Driver and let Windows search automatically. To use a driver you downloaded from the device maker, select Browse my computer for drivers instead and point to the unpacked folder.
4. Or roll back the driver. If the problem started after a driver update, select Roll Back Driver, choose a reason and select Yes. The button is greyed out when Windows has no earlier driver to return to.

   

These steps were written for Windows 11 24H2 and later. The full page with the other versions and what each result means: Update, roll back or reinstall a driver

• Read the release notes before you update again. Each driver has a PDF with *Fixed Issues* and *Open Issues*. The notes for 617.42, for example, list as fixed a crash to desktop in one game that appeared after updating to driver 616.56.
• Check that the old version stays. Look at Driver Version on the Driver tab again after a few days. If a newer version is back, the crashes can return with it.
• Tell NVIDIA. A fault in a version only gets fixed when it is reported. The address NVIDIA gives for driver feedback is [email protected].

Fix 5. Check the temperature, the power cables and the seating of the card

20 to 40 minutesLow riskCrashes come after minutes of load, the PC restarts with no dump, the build uses a riser cable, or the fixes above changed nothing.

Microsoft names insufficient cooling and insufficient power among the hardware reasons for a GPU timeout. NVIDIA's support asks you to check that the card has not come loose and that its power cables are fully inserted before you send a crash dump.

1. Press Ctrl + Shift + Esc to open Task Manager and select Performance. A dedicated graphics card shows its temperature in degrees Celsius next to its entry in the list.
2. For a running log, open Terminal and start NVIDIA's own command-line tool, which comes with the driver. It prints a table with the temperature, the fan speed and the power draw every 5 seconds. If Windows does not find the command, use Task Manager only.

   nvidia-smi -l 5

3. Start the game in windowed mode and watch the values until the crash. Note the last temperature and whether the power draw jumped just before it.
4. Shut down the PC, switch off the power supply and unplug it.
5. Open the case. Press the graphics card firmly into its slot and check that its bracket screws are tight.
6. Unplug each power cable of the card and plug it in again until it clicks, at the card and at the power supply.
7. If the card sits on a riser cable, plug it straight into the motherboard slot for the test. NVIDIA warns that riser cards can introduce stability problems.
8. Blow the dust out of the card's cooler and the case fans, then close the case and test again.

• What temperature is too high? It depends on the card, and we give no number we cannot source. What matters is the pattern: a temperature that climbs steadily to the same value before every crash points to cooling.
• The power supply. NVIDIA's troubleshooting page notes that wattage is not the whole story and that other properties of a power supply can prevent it from working with a card. A crash at the moment the load jumps is the typical sign.
• Ventilation. NVIDIA also names a PC placed in a closed desk compartment as a cause of overheating. Test with the side panel off: if that cures it, the case airflow is the problem.

Fix 6. Test the memory

30 minutes or moreNo riskCrashes also happen outside games, stop errors name different drivers, or the PC became stable in Fix 2 only after the memory profile was set to Default.

The graphics driver keeps its data in system memory like every other driver, so a bad memory cell can make nvlddmkm fail without any fault in the card. Microsoft lists memory configuration and timings, and defective memory modules, among the reasons for a display driver timeout.

The command that opens Windows Memory Diagnostic

mdsched.exe

1. Open Windows Memory Diagnostic. Press Windows + R, type mdsched.exe and select OK. You can also search for Windows Memory Diagnostic and open it from the results.
2. Choose when to test. Select Restart now and check for problems (recommended) to test right away. Select Check for problems the next time I start my computer if you want to finish your work first: the test then runs at the next start.
3. Let the test run. The PC restarts into a blue screen that shows the progress of the test. You do not have to do anything. When the test ends, the PC restarts by itself and Windows loads.
4. Look for the notification. Sign in. After a short while Windows shows a notification with the result, also when no errors were found. If you missed it, read the result in Event Viewer.
5. Open the System log. Press Windows + R, type eventvwr and select OK. In Event Viewer, expand Windows Logs and select System.
6. Find the result. Right-click System and select Find. Type MemoryDiagnostics-Results and press Enter. Select the event that is found: its text says whether the test detected errors.

These steps were written for Windows 11 25H2 and 26H2, Windows 10 22H2. The full page with the other versions and what each result means: How to test your RAM with Windows Memory Diagnostic

A test that finds errors is a clear result: remove one module at a time to find the bad one, or return the memory under warranty. A test that finds nothing is weaker evidence, because short tests miss faults that show only when the memory is warm. If the PC is stable with the memory profile at Default and unstable with XMP or EXPO, stay on Default or choose a slower profile.

Fix 7. Remove TDR registry changes instead of adding them

5 minutesLow riskYou, a tuning tool or a video guide added TdrDelay, TdrDdiDelay, TdrLevel or a similar value, or you are about to.

Many guides present a longer TDR delay as the cure. It is not one: it only changes how long Windows waits before it resets a card that has already stopped answering. Microsoft states that end users should not change these values, and a wrong value turns a recoverable flicker into a frozen PC or a blue screen.

1. Press Win + R, type regedit and press Enter. Select Yes.
2. Paste this path into the address bar of Registry Editor and press Enter: HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers
3. Select File > Export and save the key as a backup.
4. In the right pane, look for values whose names begin with Tdr.
5. Right-click each one, select Delete and confirm. Windows then uses its defaults.
6. Restart the PC.

• When a longer delay is legitimate. NVIDIA's knowledge base suggests a higher timeout for one case: a rendering job that honestly needs more than two seconds per task. Its first suggestion even there is to reduce the workload. A game that crashes is a different case.
• Why NVIDIA itself sets TdrLevel to 1. Its support uses that value for a short time to force a full memory dump for analysis, and its article tells you to remove it afterwards. It is a diagnostic step.
• What a longer delay looks like. The picture stays frozen for the whole delay before anything happens. If the card has truly hung, you only wait longer for the same reset.

If the crashes continue: what to collect and when it is the hardware

After Fixes 1 to 7 you have a PC at default speeds, with a cleanly installed current driver, checked cables, tested memory and default TDR values. If it still logs nvlddmkm bursts with crashes, software is an unlikely cause. NVIDIA's own troubleshooting page gives the deciding test: put the card into another computer.

• Signs that speak for hardware: crashes at default speeds in every 3D program, a PC that restarts with no dump, and extra lines in the events that differ each time.
• Signs that speak for software: one game only, a clear start date that matches a driver, a game patch or a Windows update, and stability after a rollback.
• Application crash dumps: %LOCALAPPDATA%\CrashDumps. NVIDIA says the file is usually named after the program, like ApplicationName.exe.1234.dmp.
• Recovered timeouts: %SystemRoot%\LiveKernelReports. Blue screens: %SystemRoot%\Minidump and C:\Windows\memory.dmp.
• What to send: the list of incidents from Fix 1 with their extra lines, the driver version, the card model, the power supply model and the dump files. Dumps are large, so NVIDIA asks for a link to a file-sharing service instead of an attachment.

How to keep nvlddmkm errors from coming back

One more habit helps with any future fault: note the date whenever you change a driver, a firmware setting or a part. The System log tells you when the first nvlddmkm burst appeared, and a dated change next to it is the shortest way to the cause.

Frequently asked questions

What is Event ID 0 in Windows, including Windows 11?

Event ID 0 is not a Windows error code. Every program that writes to the event log numbers its own events, and Microsoft's documentation says the number only has a meaning together with the source. A program that does not number its events writes all of them with 0. Windows 11 and Windows 10 behave the same here. To learn what an Event ID 0 entry is about, look at the Source column and at the text after "The following information was included with the event". The number tells you nothing by itself.

Is my NVIDIA driver really not installed or corrupted?

No, the sentence does not prove that. Event Viewer shows it for any event whose text it cannot find, whatever the source. In the nvlddmkm entries the closing line usually reads "The message resource is present but the message was not found in the message table". That is Windows error 15027 and it means the driver's message file was found and opened. Only this one number has no sentence in it. A driver that is truly missing shows up elsewhere: the Driver tab of the card in Device Manager then names no NVIDIA driver, and games do not start at all.

What do Event ID 13, 14 and 153 from source nvlddmkm mean?

NVIDIA does not publish their meaning, so only the extra line can be read. In logs that users post, event 13 carries a "Graphics Exception" line, event 14 carries register names with hexadecimal values, and event 153 carries "Error occurred on GPUID: 100" or "Reset TDR occurred on GPUID:100". All of them are reported together with a driver timeout: a black screen for a moment, a closed game, or a restart. They are handled exactly like Event ID 0 from the same source: find the incident, then work through the fixes on this page.

Will sfc /scannow or DISM fix this error?

Almost never. System File Checker and DISM repair files that belong to Windows. The NVIDIA driver is not one of them, and the missing description is not a damaged file in the first place. The old version of this guide listed the system file check as the first method, and we removed it for that reason. It does no harm, and it is a reasonable step when the PC also shows faults that have nothing to do with graphics. If you want to run it anyway, the steps are here: Run DISM and SFC to repair Windows system files.

Does increasing TdrDelay fix nvlddmkm crashes?

No. TdrDelay is the number of seconds Windows waits before it declares the graphics card frozen. The default is 2. A higher value does not make the card answer: it makes Windows wait longer in front of a frozen picture before the same reset. Microsoft writes that these registry values are for driver development and that end users should not change them. NVIDIA suggests a longer timeout only for heavy rendering work that needs more than two seconds per task. If you already added the value, Fix 7 shows how to remove it.

What is LiveKernelEvent 141, and is it a blue screen?

It is not a blue screen. LiveKernelEvent 141 is the name Reliability Monitor gives to a live dump with the code 0x141, VIDEO_ENGINE_TIMEOUT_DETECTED. Microsoft says this code is never used for a real stop error and only identifies live dumps: one engine of the graphics card did not answer in time, Windows reset it and saved a report while the system kept running. Code 117 is the same for the whole adapter. The files are in the LiveKernelReports folder inside the Windows folder. Many of them in a short time lead to the real blue screen, VIDEO_TDR_FAILURE.

Does this error mean my graphics card is dying?

Not by itself. The same events come from an overclock that is slightly too high, a memory profile, one faulty driver version, dust, a loose power cable or a weak power supply. A failing card becomes likely only when the crashes continue at default speeds, with a cleanly installed driver, in every 3D program, and above all when the card also fails in another PC. That swap is the test NVIDIA itself recommends. If the card passes in another PC, look at the power supply, the memory and the motherboard of your own.

I see Event ID 0 from another source, such as gupdate or OVRServiceLauncher. Is it the same problem?

It is the same Windows behaviour and a different program. Any source that logs without a registered message text gets the sentence "The description for Event ID 0 from source [name] cannot be found". It has nothing to do with NVIDIA or with graphics. Read the text after "The following information was included with the event": that is what the program wanted to say, often a plain status line and no error at all. This page covers the nvlddmkm case only. If the entry has the level Information and nothing misbehaves, you can ignore it.

Can I make Event Viewer show the real description?

No. The sentence for each event number lives in a message file that the program's publisher ships, and the nvlddmkm entries end with the note that the message was not found in that file's table. Nothing on your PC is missing that you could restore. A new driver version could add the text, but we found no NVIDIA statement about it. You lose little: the values the driver logged are shown in full after "The following information was included with the event", and on the Details tab of the event.