The short answer
The absence of a blue screen is the most useful clue you have. A blue screen means Windows detected a problem and stopped deliberately. An instant restart with no screen at all means the machine lost power or was reset at the hardware level before Windows could react — which points at the power supply, overheating protection, or a loose connection, not at drivers or software. Start by checking Event Viewer for a Kernel-Power event, then test temperatures under load.
There are two very different failures that both get described as “my PC keeps restarting”, and telling them apart is most of the work.
If you get a blue screen, even for a fraction of a second, Windows caught something — a driver fault, a memory error, a filesystem problem — and chose to stop. That leaves evidence: a minidump file, a stop code, an event log entry with detail.
If the machine simply goes black and reboots, with no screen and no sound, Windows never got the chance. Something below the operating system pulled the plug. That is a much shorter list of suspects and it is worth working through in order.
What causes an instant, silent restart
Power delivery. A power supply that cannot hold up under a transient load will drop its rails for a few milliseconds, and the motherboard resets. This is why the problem so often appears during gaming, rendering or any sudden spike in demand, and never at idle. Power supplies also degrade with age — capacitors lose capacity over years — so a unit that was adequate when new may not be now.
Thermal protection. Modern CPUs throttle long before they shut down, so a CPU thermal event usually means something quite wrong: a heatsink that has come loose, pump failure on a liquid cooler, or dried-out thermal paste. Motherboard VRM overheating and GPU thermal protection produce the same instant cut, and VRM temperatures are rarely monitored by default.
Loose or failing connections. The 24-pin motherboard connector and the 8-pin CPU power connector both need to be fully seated. A connector that is almost home works fine until current rises. The same applies to GPU power cables, and to the increasingly common problem of a cable that is seated at the PSU end but not quite at the card.
Memory faults. Bad RAM more often causes blue screens, but a severe fault, or a module that is not making proper contact in its slot, can cause a hard reset instead.
External power. A failing wall socket, an overloaded extension lead, or a UPS with a dying battery can all cause brief interruptions your PC notices and you do not. A house with lights that dim when the washing machine starts is worth taking seriously here.
Before anything else, open Event Viewer and check. Press Win + X, choose Event Viewer, expand Windows Logs and click System. Look for Kernel-Power, Event ID 41 at the time of each restart. That event means exactly one thing: the system shut down without going through the normal process. It confirms you are dealing with a hardware-level reset rather than a software crash.
Confirm what kind of restart you are having
Open Event Viewer as above and note the event IDs around each restart. Kernel-Power 41 with no preceding error means a hard reset. A Bugcheck event means Windows did crash and you have a blue screen you are not seeing.
If you find Bugcheck entries, turn off automatic restart so you can read the stop code: System Properties > Advanced > Startup and Recovery > Settings, and untick Automatically restart. Then the next crash will show you a blue screen with a code worth searching.
If it is purely Kernel-Power 41, continue with the hardware steps below and stop troubleshooting drivers.
Watch temperatures under real load
Install a monitoring tool that logs to file rather than just displaying, so you have data covering the moment of the restart.
Run a sustained load — a game, a render, or a stress test — and watch CPU package temperature, GPU temperature and, if your board exposes it, VRM or motherboard temperature. A restart that coincides with a temperature spike above the mid-nineties Celsius is thermal.
If temperatures climb unusually fast from idle, suspect heatsink mounting. If a liquid cooler’s pump speed reads zero or fluctuates, the pump is failing. If the CPU is fine but the machine still cuts out under GPU load, look at the graphics card and its power delivery instead.
Reseat every power connection
Shut down, switch off at the wall, and hold the case power button for ten seconds to discharge.
Unplug and firmly reseat the 24-pin motherboard connector, the 4 or 8-pin CPU power connector at the top of the board, and every PCIe power cable at both the graphics card and the power supply. Listen for the click on each.
If you are using a modular power supply, confirm every cable is one that came with that specific unit. Modular cables are not interchangeable between brands or even between models from the same brand, and using the wrong one is a genuine hazard as well as a cause of instability.
Avoid daisy-chaining a single PCIe cable to feed two connectors on a power-hungry card. Use separate cables where the card has multiple inputs.
Test the power supply by substitution
There is no reliable software test for a PSU. The practical approach is to reduce the load and see whether stability returns.
Remove the graphics card and run on integrated graphics for a few days if your CPU supports it. If the restarts stop entirely, the PSU cannot support the card — either because it is underpowered, degraded, or faulty.
Alternatively, cap the frame rate or lower the power limit on the GPU to reduce peak draw. If a modest reduction in load removes the problem completely, that is a strong signal.
Borrowing a known-good power supply of adequate capacity is the definitive test. If you are buying, factor in headroom for transient spikes — the average draw is not what trips a marginal unit.
Test the memory
Run a proper memory test overnight rather than the quick Windows one. Boot from a memory testing tool and let it complete several full passes.
Also try reseating the modules, and test with one stick at a time in the primary slot. A module that is not fully clipped in at both ends is surprisingly common after any work inside the case.
If you are running an XMP or EXPO memory profile, turn it off in BIOS and run at the default speed for a day. Memory overclocking profiles are a frequent cause of instability that people do not think of as overclocking because it is a single switch.
Rule out the wall
Plug the PC directly into a wall socket rather than an extension lead or a power strip shared with high-draw appliances. Try a different circuit if you can.
If you use a UPS, check its battery health. A UPS with a failing battery can pass through brief interruptions rather than covering them, which is worse than no UPS at all.
A telltale sign of supply-side trouble is that other devices in the room misbehave at the same moments — lights dimming, a monitor blinking, a router rebooting. If the PC is not alone in its complaints, the problem is not in the PC.
Reset the BIOS and undo any overclocks
Load optimised defaults in BIOS. This clears any overclock, any voltage adjustment, and any memory profile in one step.
Run the machine at stock settings for several days. If it is stable, reintroduce changes one at a time. Many restart problems trace back to a mild overclock that was stable when it was set and is no longer stable as the hardware has aged or as ambient temperatures have changed.
While you are in BIOS, check that the CPU fan header reports a sensible speed and that no fan is reading zero.
Reading the pattern
Only under heavy load, never at idle. Power supply or thermal. Test temperatures first, then reduce load to confirm.
At idle, seemingly at random. More likely a loose connection, a failing PSU at low load, or mains supply. Reseat everything.
Always after a similar period of uptime. Thermal buildup, often VRM or a clogged heatsink. Check dust and airflow.
Started after adding a new component. The new component’s power draw pushed a marginal PSU over the edge, or a cable was disturbed during installation.
Worse in warm weather. Thermal, or a PSU losing capacity at higher ambient temperature. Both are real.
Accompanied by a click from the PSU. The protection circuit is tripping. That is the PSU telling you directly.
Blue screen that you catch on camera. Not this problem at all — you have a software or driver fault, and the stop code will tell you which.
Do not keep running a machine that restarts under load while you work through this slowly. Unclean shutdowns risk filesystem corruption, and repeated protection trips are hard on the components involved. Back up anything important before you start testing.
The dust factor
It is worth a separate mention because it is so often the whole answer. A machine that has been running for two or three years in a carpeted room can have heatsink fins almost completely bridged with dust. Airflow drops, temperatures climb, and a PC that was fine becomes one that cuts out under load.
Power down, take it outside, and clean the CPU heatsink, the graphics card heatsink, every fan and every intake filter. Hold fans still while you blow them out so they do not spin up and generate voltage. A dust-clogged PSU is worth cleaning from the outside but not worth opening — there are capacitors inside that hold charge.
If temperatures drop substantially after cleaning and the restarts stop, put a reminder in your calendar to do it twice a year. It is the highest-value maintenance there is.
Why is there no blue screen if the PC is crashing?
Because it is not crashing in the software sense. A blue screen is Windows deliberately stopping after detecting a fault. An instant restart means power was lost or the board was reset before Windows could respond, which is a hardware-level event.
Can a bad graphics driver cause restarts with no blue screen?
Rarely. A driver fault normally produces either a blue screen or a recoverable display driver reset. If you see Kernel-Power 41 with no other errors, drivers are an unlikely explanation.
How do I know if my power supply is too weak?
Peak transient draw matters more than the average figure on a spec sheet. A practical test is to reduce load — remove the graphics card or cap frame rates — and see whether stability returns. If it does, the supply cannot handle the peaks.
Does a restart mean my CPU is damaged?
Almost never. Thermal shutdown is a protection mechanism that exists precisely to prevent damage. It indicates a cooling problem, not a broken processor.
Should I replace the power supply just to rule it out?
Borrow one first if you possibly can. A PSU is one of the few components where buying to test is expensive and where the symptom overlaps heavily with a loose cable, which costs nothing to check.
The bottom line
Check Event Viewer first. Kernel-Power 41 with nothing before it tells you to stop looking at software entirely, and that single fact saves most people hours.
From there, work through heat, connections and power in that order: log temperatures under real load, reseat every power cable at both ends, clean the dust out, and reduce the load to see whether stability returns. Reset the BIOS to defaults to eliminate any overclock or memory profile. If the machine is stable at stock with a lighter load and unstable otherwise, the power supply is the prime suspect and substitution is the only way to be sure.
