Start your 3-day free trial
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.


If a Windows blue screen keeps returning, preserve the stop code and identify what changed before trying broad repairs. Repeated bug checks are usually caused by a faulty or incompatible driver, unstable hardware, corrupted system data, or a problem that appears under a repeatable workload.
Back up important files as soon as Windows stays usable long enough. Stop testing if the computer has battery swelling, liquid damage, a burning smell, smoke, electrical noise, or severe heat.
Key Takeaways
- Photograph the stop code, failed module, progress percentage, and time.
- Separate one application crash from an operating-system bug check.
- Remove one recent peripheral or change at a time and keep a test log.
- Use Safe Mode and official drivers to narrow the cause, not to hide it.
- Back up before uninstalling major drivers, restoring Windows, or reinstalling it.
Start with the device and app troubleshooting guide if you need a general method for changing one variable at a time.
A blue screen is a Windows stop error, not merely a blue application window, frozen desktop, or game crash. Microsoft describes it as an unexpected restart or shutdown caused by a serious problem and recommends noting the stop code and any named module.[1]
Use a phone camera if the screen disappears quickly. Save:
Do not delete crash dumps or run a cleaner before support has a chance to use them. Do not assume a filename on the screen proves that one driver is guilty; it may be the component that detected corruption caused elsewhere.
If Windows starts, open Reliability Monitor or the system’s event history and match the failure time. Record the bug-check entry and preceding hardware or driver errors without treating every warning as causal. If Windows never reaches the desktop, move to the recovery and Safe Mode branch.
Create a short timeline for the last three failures. A repeated trigger is more useful than a long list of installed software.
| Trigger pattern | First branch to test |
|---|---|
| Immediately after connecting a device | Peripheral, cable, dock, port, or its driver |
| During graphics-heavy work | Display driver, GPU temperature, power, or hardware |
| During large file access | Storage device, controller, free space, or file-system health |
| On sleep or wake | Firmware, chipset, graphics, storage, or power management |
| Soon after an update | Update completion, official driver compatibility, or rollback path |
| At random workloads | Memory, storage, heat, power, firmware, or a low-level driver |
Disconnect newly added nonessential peripherals while the computer is shut down. Leave only display, keyboard, mouse, and required power. Test one known workload once; repeated crashes can worsen file corruption and are not a stress test.
If a single application closes but Windows remains responsive, use the app launch and crash guide. If the whole system shows a stop code and restarts, keep following the operating-system path.
For a crash that began during a Windows update, first verify whether the update completed. The Windows Update guide covers download, installation, restart, and recovery stages without turning every blue screen into an update problem.
List changes from the day before the first crash through the first confirmed failure. Include Windows updates, manufacturer driver packages, BIOS or UEFI updates, security software, virtualization tools, storage utilities, new RAM, a new GPU, a dock, and external storage.
Reverse only one reversible change at a time. Remove a newly connected peripheral, then test. If the computer is stable, reconnect it through a different known-good cable or supported port before deciding the device is faulty.
Use Windows Update and the computer or component manufacturer’s support page for drivers. Microsoft’s blue-screen guidance recommends removing new hardware and, when Windows can start, using Device Manager to update, disable, or uninstall a device associated with the problem.[1] Do not install “driver updater” bundles or drivers selected only by a similar product name.
If a newly updated driver clearly matches the first crash, use the supported roll-back option when available. Preserve the installed and previous version numbers. Do not disable driver-signature enforcement to make an unknown package load; that removes a protection while introducing another low-level variable.
Firmware rollback can be risky and may be blocked by the manufacturer. Do not force it from a forum procedure. Check model-specific advisories and maintain stable power during any approved firmware operation.
Safe Mode starts Windows with a limited set of files and drivers. Microsoft includes it in Windows Startup Settings and distinguishes ordinary Safe Mode from Safe Mode with Networking.[2] Start without networking unless you need an official download and understand the additional services it enables.
If Windows fails to start normally, reach Windows Recovery Environment through the supported recovery path, then choose Troubleshoot, Advanced options, Startup Settings, Restart, and the appropriate Safe Mode option. Menu wording can vary by Windows release and device recovery configuration.
Once in Safe Mode:
Safe Mode stability does not prove the hardware is healthy. It shows that the failure may depend on a driver, service, startup component, resolution, power state, or workload absent from Safe Mode. A crash in Safe Mode raises the priority of hardware, storage, core system data, and firmware checks.
Do not leave the system permanently in a reduced boot configuration merely because crashes disappear. The goal is to identify and remove the cause, then return to a normal verified startup.
Start with free space, official updates, and Device Manager status. A nearly full system volume can prevent updates and dump creation, but high disk usage is a different symptom; use the 100% disk usage guide rather than applying unrelated registry tweaks.
Run Windows Memory Diagnostic when crashes are random, memory-related, or began after RAM work. Save the result. For manufacturer systems, use the approved preboot memory and storage diagnostics as well, because they can test without the normal Windows driver stack.
Use system-file and storage checks only from Microsoft or the device maker’s current instructions. Keep the computer on stable power, do not interrupt a repair, and back up before a tool that may modify file-system metadata. A storage health warning, repeated read errors, or a failed manufacturer diagnostic is a replacement-and-recovery signal, not a reason to loop repairs.
Check temperatures and airflow without opening a device that is not user-serviceable. Remove dust only by the documented maintenance method. If the laptop’s fan behavior is the main clue, collect that evidence separately rather than blocking vents or forcing maximum fan utilities.
Avoid third-party “registry cleaners,” dump cleaners, overclocking profiles, and simultaneous stress tools. They erase evidence or add instability. Restore CPU, GPU, and memory tuning to manufacturer defaults before interpreting diagnostics.
Microsoft’s recovery guidance maps different failures to Startup Repair, uninstalling an update, System Restore, reset, installation media, or a recovery drive.[3] Choose the least destructive option that matches the evidence.
Use this order:
Do not repeatedly interrupt startup to force recovery if Windows already offers a supported entry. On an encrypted device, confirm the recovery key before changing boot, firmware, or storage settings. A work-managed computer may require IT because security agents, disk encryption, and policy are part of the startup chain.
If Windows starts only briefly, prioritize backup and evidence over optional updates. If it never starts and hardware diagnostics fail, recover data through a qualified service rather than repeatedly writing to a failing drive.
Escalate when the same crash returns after the matching recent change is removed, Windows crashes in Safe Mode, memory or storage diagnostics fail, or the computer has impact, liquid, heat, power, or battery damage. Provide the stop codes, timestamps, driver versions, diagnostic codes, and a concise trigger timeline.
Do not replace multiple parts at once. If removable RAM is explicitly user-serviceable, follow the exact model’s electrostatic-discharge and seating procedure; otherwise leave the enclosure closed. Soldered memory, internal batteries, liquid exposure, and power faults need qualified service.
Slow startup without a stop error belongs in the Windows startup guide. Keeping these symptoms separate prevents an aggressive recovery step from being applied to the wrong failure.
A repeatable code can point to the same driver, hardware path, or corruption, but it is not a verdict by itself. Compare the trigger, failed module, recent changes, dumps, and diagnostics.
Changing codes can occur with unstable memory, storage corruption, overheating, power faults, or a driver damaging unrelated data. Restore tuning defaults and run official memory and storage diagnostics.
Save work and back up first. One isolated stop may not recur, but a second failure or a hardware warning justifies a structured investigation rather than normal use.
Safe Mode is mainly a diagnostic environment. Stability there helps isolate optional drivers and services, but you still need to identify the cause and return to normal startup.
Not before diagnosis. Dumps can preserve the stop code and module context needed by support; a cleanup tool may erase the best evidence.
An update, an accompanying driver, or an incomplete installation can be related when the timing matches. Confirm the update history and use Microsoft’s supported uninstall or recovery option instead of guessing.
No as a routine fix. It weakens a security boundary and may allow an untrusted or incompatible kernel driver to create more crashes.
Reinstall only after backup, hardware diagnostics, targeted driver checks, and less destructive recovery options. A reinstall will not repair failing RAM, storage, power, or cooling hardware.
Sources checked 4 September 2026.
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.