What Chen actually described
Chen says that holding a PC's power button for 12 seconds turns it off at the hardware level without notifying the operating system. He calls this a safety feature for when the "Shut down" command isn't working. According to The Register's account, after eight seconds the Windows kernel realizes a forced power-off is under way and tries to take a kernel dump for diagnostics. That dump can be uploaded to Windows Error Reporting the next time you start the system.
Chen also notes the weak point. Sometimes four seconds isn't enough time to gather enough information, and the kernel dump is incomplete. He believes Windows Insider builds show a message asking you to let go of the power button, which gives the kernel more time to finish. He says "I believe", so treat that as unconfirmed.
Why "LPBH" and "black screen" matter
Microsoft internally classifies these emergency dumps as "long power button hold" (LPBH) failures. One recorded detail is whether the screen consists entirely of black pixels. Those cases are sub-classified as "black screen LPBH" and go to a special virtual team for analysis. Chen says he is a member of it.
Chen lists the possible causes of a fully black screen:
- A broken video driver.
- A working driver that is being handed black pixels by the desktop compositor.
- Nothing drawing to the screen at all, sometimes because the process that normally draws is hung.
These are categories for triage. They are not a diagnosis for your particular machine. A black screen also isn't always a Windows fault. Chen's Windows 3.1 anecdote involves a user who set every color in the Colors control panel to black.
The timing is not universal
Treat the eight- and 12-second figures as Chen's description of the general design, not a guarantee for every PC.
- Microsoft's power-button bugcheck documentation: it describes a firmware hard power-off timeout that is "usually about 10 seconds". It says the timer is shorter on some systems, such as ARM64 laptops.
- Event ID 41 guidance: it mentions forced shutdowns after holding the button for at least four seconds. That is a different matter from when a dump is captured.
- Chen's own comment section: readers report desktops that cut power after about five seconds. In that case Windows would never reach the eight-second mark, so no dump could be taken. This is anecdotal, but it fits the logic of the mechanism.
- Older Microsoft context: an earlier Chen post says Windows recognizes a long power button hold at seven seconds, and the UEFI forces a power-off at ten seconds. That differs slightly from the eight-second figure in the new post, which is more evidence that exact numbers vary by implementation or era.
A related feature you can configure
Microsoft documents a separate, opt-in mechanism: Bug Check 0x1C8, MANUALLY_INITIATED_POWER_BUTTON_HOLD. It is not the same thing as the emergency behavior Chen describes, and Microsoft hasn't tied the two together.
- It is enabled by setting the DWORD
PowerButtonBugcheckto 1 underHKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Power. - The bugcheck fires when the button is held for 7 seconds but released before the UEFI reset at 10 seconds.
- You then see a black screen reading "Please release the power button. We just need a few more seconds to shut down", with a percentage indicator.
- If you keep holding, the hard power-off can interrupt the dump and corrupt it.
- It is available in Windows 10 1809 / Windows Server 2019 and newer. The device needs a GPIO-based power button, firmware that routes the power event to the Windows Power Manager, and the registry feature enabled.
- Microsoft also documents a live-dump variant, 0x11C8, controlled by a
PowerButtonLiveDumpvalue. If both are enabled, the bugcheck takes priority.
Chen's emergency dump is described as happening during a forced power-off. The 0x1C8 and 0x11C8 features are explicit configuration options. Microsoft's documentation doesn't say the registry value is a prerequisite for Chen's behavior, so don't assume either way.
Does Microsoft automatically receive the dump?
Not necessarily. Several things need to be kept apart.
- The event versus the dump. Microsoft's required-diagnostics documentation lists a
LongPowerButtonPressDetectedfield, which indicates whether the user was holding the power button. That tells Microsoft a long press happened. It is not a memory dump. - Where dumps sit in the settings. Microsoft's guidance for organizations says Required diagnostic data is a limited set covering device configuration, quality metrics and compatibility data. Optional diagnostic data adds enhanced error reporting, including the memory state of the device when a crash occurs. Microsoft warns that this memory state may unintentionally include user content, such as parts of a file you were using.
- What Microsoft hasn't said. The available documentation doesn't spell out exactly what an LPBH dump contains, so it would be wrong to claim that every one holds personal file content, or that every one is uploaded.
- Admin controls. Microsoft documents a Group Policy that limits dump types. When enabled, Windows Error Reporting sends only kernel mini dumps and user-mode triage dumps. It is found under Computer Configuration > Administrative Templates > Windows Components > Data Collection and Preview Builds. This is general dump policy, not an LPBH-specific guarantee.
Checking your own setting
On Windows 11, open Settings > Privacy & security > Diagnostics & feedback and expand Diagnostic data. Windows Latest reports that turning off "Send optional diagnostic data" removes the main permission Microsoft names for collecting crash dumps. The same report says there is no separate switch for power-button dumps. Turning this off does not stop required diagnostic events from being recorded.
What Event ID 41 can and can't tell you
After a forced shutdown, you can look in Event Viewer under Windows Logs > System for Kernel-Power Event ID 41. Microsoft says a press-and-hold shutdown logs a nonzero PowerButtonTimestamp. Some important limits apply:
- Event ID 41 means Windows detected that the previous shutdown wasn't clean. It does not prove a dump was written or sent.
- Microsoft documents cases where
BugcheckCodeandPowerButtonTimestampare both zero. Windows may be unable to write error data to disk, or the machine shut down too quickly to record anything. - Microsoft recommends the power-button method only when there is no alternative. Normal shutdown closes files and notifies services and applications so they can write unsaved data and flush caches.
- If Event ID 41 shows all zero values, Microsoft also suggests checking for an Event ID 46 from volmgr ("Crash dump initialization failed!"). That can point to a missing or misconfigured dump file, which by default is the pagefile.
What this means in practice
For most users, nothing changes. If your PC is genuinely frozen, holding the button is still the right last resort. What's useful is knowing that:
- A long hold on a hung machine may produce diagnostic data, but it's an opportunistic capture, not a promise.
- A recorded long-press event and an uploaded memory dump are different things governed by different diagnostic settings.
- If you want to limit what leaves the machine, the Optional diagnostic data toggle is the control Microsoft points to. Administrators can use the dump-type policy.
- If you troubleshoot recurring freezes, treat Event ID 41 as a clue and not a verdict. Check the zero-value cases and the dump configuration.
Chen's post doesn't say how often these dumps lead to fixes. It does show that the "black screen of nothing" is a category Microsoft engineers actively sort and analyze. The next time your screen goes dark and you reach for the power button, at least there's a reasonable chance someone with a debugger will be interested in the result.
References
- Forcing a System Crash with the Power Button - Windows drivers | Microsoft Learn learn.microsoft.com
- Why are the Windows chassis hardware button hotkeys so strange? - The Old New Thing devblogs.microsoft.com
- The support case of a customer who said that all they could see was a black screen - The Old New Thing devblogs.microsoft.com