Display, Event 4101
"Display driver nvlddmkm stopped responding and has successfully recovered." The operating system reset the GPU after a timeout.
The game freezes. The screen goes black, sometimes for a second, sometimes long enough to think the PC has died. Then the desktop comes back, the game has crashed or shows a black window, and a small notification says Display driver nvlddmkm stopped responding and has successfully recovered. In Event Viewer there is an error from source nvlddmkm with Event ID 153, and when you open it, Windows says the description for the event cannot be found.
That missing description is not your PC being broken in an extra way. NVIDIA does not publish text for that event. The useful facts are elsewhere: what Windows itself recorded, what the machine was doing in the minute before, and whether the reset follows load, heat, idle, one game or one driver version.
nvlddmkm, Event 153 (often with 13 or 14)
The driver's own entry for the same moment. Undocumented publicly. Treat it as a timestamp, not a diagnosis.
Stop 0x117 or 0x116
Too many resets in a minute, or a driver that cannot be reset, becomes a blue screen instead of a flicker.
In this guide
What actually happened: a two second timeout
Windows has a watchdog for the graphics card called Timeout Detection and Recovery (TDR). The GPU scheduler in the DirectX graphics kernel hands the card work. If the card takes too long, the scheduler asks it to stop and switch (preempt). If the card can neither finish nor preempt within the timeout, Windows decides the GPU is frozen, tells the driver to reset the hardware, purges video memory and redraws the desktop. Microsoft documents that the only visible sign of a successful recovery is a screen flicker and that message.
2 sdefault timeout before the GPU is treated as hung (TdrDelay)
5 in 60 sresets Windows allows within a minute (TdrLimitCount, TdrLimitTime)
6thhang within that minute stops the PC with 0x117
5 sfor threads to leave the driver before 0x116 (TdrDdiDelay)
So the event is a symptom with a stopwatch in it. Something kept the GPU busy, stalled or unresponsive for more than two seconds. A game that runs a heavy frame is supposed to finish in milliseconds. Anything that holds the card for two full seconds is a fault somewhere in the chain: the driver, the card's stability at its current clocks and voltage, its temperature, its power, the PCI Express link, or the rest of the system feeding it.
Write down what you saw. Did only the screen go black, or did audio stop too? Did the game crash
to the desktop or keep running with a black window? Did it happen in a game, at idle, while watching a
video or when switching windows? Was a monitor waking up? These details choose the next test better
than the event number does.
Read the pattern, not the event ID
Open Reliability Monitor with perfmon /rel and find the days with the
reset. It puts driver installs, Windows updates and application crashes on one timeline, which makes a
repeated neighbour easy to spot. Then use the table.
| When it happens | Moves up the list | What it does not prove |
|---|---|---|
| Only under heavy game load, after minutes of play | Heat, GPU core or memory overclock, undervolt, power delivery | That the card is dead |
| At the desktop, while browsing, on alt-tab, on monitor wake | Power state transitions, the driver version, display cable or adapter | That the GPU cannot handle load |
| Started the day a driver or Windows update installed | That driver version; try the previous one | That the newest driver is always the fix |
| Only in one game or one engine | That game's settings, its shader cache, its overlay, a known issue in its patch notes | A system-wide hardware fault |
| Stable at stock, unstable with Afterburner or a factory OC profile | Lost stability margin at those clocks | That the defaults are "slow" |
| Also whole-PC freezes or restarts, WHEA warnings, RAM overclock | System instability outside the GPU: memory, CPU, board. See the WHEA guide | That NVIDIA's driver is the culprit |
| Many resets per session, getting worse over weeks | Hardware degradation, heat from dried paste or pads, a failing power connector | Anything until tested at defaults |
A safe isolation order
- Keep the evidence. Note the exact time, the pattern above, and the events around Event 4101 in Windows Logs → System. If a blue screen happened, note 0x116 or 0x117.
- Return everything to default. Remove GPU core and memory overclocks, custom curves and undervolts. Test the card at its reference clocks if the maker's tool allows it. Also return RAM XMP or EXPO and CPU tuning to default for the test: a GPU reset can be the first symptom of an unstable system. The RAM stability guide shows a controlled A/B test.
- Reinstall the driver cleanly. Use the clean installation option in NVIDIA's installer. If the problem survives, a full removal in Safe Mode with Display Driver Uninstaller is the stronger version of the same test. Then try the previous stable driver as well as the newest: both directions are useful data.
- Log temperature and clocks while reproducing. Watch GPU temperature, hotspot if your tools show it, clocks and power draw in the few minutes before a reset. Clocks falling as the temperature climbs point at cooling. A reset at normal temperature points elsewhere. Stop if anything smells hot or behaves dangerously.
- Check power and the link physically. With the PC switched off and unplugged, reseat the GPU and its power connectors. A 12V-2x6 or 12VHPWR plug must be fully inserted with no gap. Use separate PCIe cables from the power supply where the card maker asks for it, and avoid extensions and risers during the test.
- For idle and desktop flickers, set Power management mode to Prefer maximum performance in the NVIDIA Control Panel as a diagnostic, and try another cable or port, especially through adapters. If the flickers stop, you have learned it is a power state or display path problem. Set it back afterwards if you do not want the extra idle power.
- Change one thing at a time, and give each change the same session that used to fail. Three changes at once can fix it and teach you nothing.
If resets continue at full default settings with a clean driver, normal temperatures and correctly
seated power, test the card in another PC or another card in yours before buying anything. That
is the step that separates the card from the rest of the machine.
Why raising TdrDelay is the wrong fix
The most copied fix on the internet is a registry value: create TdrDelay under
HKLM\System\CurrentControlSet\Control\GraphicsDrivers and set it to 8 or 10 seconds.
Microsoft documents that key and its neighbours as settings for testing and debugging during driver
development, and states directly that end users should not change them.
The reason is simple. The delay is the watchdog, not the problem. With a longer delay the GPU still
stalls. Windows just waits longer before recovering, so a two second flicker becomes an eight second
frozen screen, and the event that tells you something is wrong appears less often. Setting
TdrLevel to 0 turns detection off entirely, which can leave the whole desktop hung
instead of recovering. There are narrow professional cases, such as long GPU compute jobs on a
display card, where developers change these values deliberately. A game crashing at random is not
one of them.
Where PC Workman helps, and where it cannot
PC Workman keeps a local history of GPU and CPU load, temperatures, clocks and the processes that were running. After a reset you can look at the minutes before it and see whether the card was hot, whether clocks were falling, or whether it happened at idle. That is often the difference between a cooling problem and a driver or power state problem.
PC Workman does not read NVIDIA's internal driver logs, decode Xid errors or crash dumps, and cannot tell a driver bug from a failing card on its own. It is one layer next to Event Viewer, Reliability Monitor, a clean driver install and a test at default settings.
Short answers
What does nvlddmkm Event 153 mean?
It is the NVIDIA driver's own record of a GPU hang and driver reset. NVIDIA does not publish a description, so Event Viewer says it cannot find one. Windows records the same reset as Display Event 4101.
Is it a hardware failure?
Not by itself. It records a timeout. Driver problems, unstable tuning, heat, power delivery and system instability all produce it. Test at defaults with a clean driver before blaming the card.
Should I change TdrDelay?
No. Microsoft says end users should not change the TDR keys. A longer delay hides the symptom and makes each hang last longer.
Why do I get it at the desktop, not in games?
Resets at idle often follow power state changes or the display path. Test with Prefer maximum performance, another cable and the previous driver.
When does it turn into a blue screen?
On the sixth hang within a minute Windows stops with 0x117. If the driver cannot be reset within five seconds, it stops with 0x116.
Primary sources
- Microsoft: timeout detection and recovery (TDR) in WDDM.
- Microsoft: TDR registry keys, for testing and debugging only.
- Microsoft: bug check 0x117 VIDEO_TDR_TIMEOUT_DETECTED.
- Microsoft: bug check 0x116 VIDEO_TDR_FAILURE.
- NVIDIA: Xid errors, the driver's own GPU error reports.
See the minutes before the reset. PC Workman keeps a local history of GPU temperature, load and
clocks, so a black flicker is not reduced to one event without a description.
Download for Windows or get it from the
Microsoft Store.
If the whole PC restarts instead of only the screen, read Kernel-Power 41 →