PC keeps restarting out of nowhere — or worse, cutting to black mid-game like someone pulled the plug? Random restarts and shutdowns are among the most frustrating faults a computer can develop, because they’re intermittent: the machine works perfectly right up until it doesn’t. The good news is that the way it dies tells you a lot about why, and most causes can be narrowed down at home with free tools. Here’s the same diagnostic order we use at the bench.
First: make Windows tell you the truth
By default, Windows restarts the instant it crashes — which means the blue screen that would have named the fault flashes past in a fraction of a second. Before anything else, turn that off:
Press Win + R, type sysdm.cpl, press Enter.
Advanced tab → Startup and Recovery → Settings.
Untick “Automatically restart” under System failure. OK out.
Next crash, note the stop code on the blue screen — that’s your first real evidence.
Then check what’s already on record: search Reliability Monitor in the Start menu for a day-by-day crash timeline, or Event Viewer → Windows Logs → System. Event ID 41 (“the system rebooted without cleanly shutting down”) means power was lost — a hardware-leaning clue.
Restart vs hard shut-off — the clue most people miss
Windows crashed and rebooted. Leans software-side: drivers, Windows updates, or unstable RAM. You’ll usually find stop codes in Reliability Monitor.
Hardware protection pulled the plug. Leans thermal or power: an overheating CPU/GPU or a PSU tripping its protection. Often no error logged at all — just Event 41.
One edge case before the causes: if the PC keeps restarting in a loop before Windows even loads, that’s a different beast — usually failed memory training or a BIOS-level fault. Jump to our boot troubleshooting guide for that pattern; this guide covers machines that run, then die.
Cause 1: overheating
Silicon protects itself: hit the thermal limit and the system shuts down hard, typically minutes into a game or render rather than at idle. Install HWiNFO (it’s on our free downloads page) and watch CPU and GPU temperatures under load. Sustained trips into the high 90s°C on the CPU, or a GPU hot spot pushing 105°C+, and you’ve likely found your culprit.
The usual suspects, in order: dust (heatsinks and filters packed solid), a cooler that’s lost contact (dried paste after 4–5 years, or a mounting pin that popped), and case airflow (fans dead, or intake/exhaust fighting each other). A proper clean and repaste fixes a remarkable share of “dying” PCs.
Cause 2: the power supply
If the machine cuts out under load with healthy temperatures, suspect the PSU. Two things get older units: capacitors age and lose capacity, and modern graphics cards draw sharp millisecond power spikes well above their rated draw — enough to trip the over-current protection on a PSU that was “fine” with your last card. The result is an instant, silent power cut, usually at the most demanding moment.
Quick sanity check: run your build through our free PSU calculator. If your unit’s wattage is at or below the recommendation — or it’s approaching a decade old — it’s a prime suspect regardless of what the sticker says.
Cause 3: unstable RAM
Memory faults produce the most random-looking symptoms of all: restarts with different stop codes each time, crashes in unrelated programs, corrupted files. A special 2026 note: XMP/EXPO profiles — the one-click RAM overclock — are a very common cause of instability, especially on DDR5 systems. Try disabling the profile in the BIOS; if the restarts stop, you’ve found it (run the sticks at a slightly lower speed, or test each stick).
For a proper verdict, run MemTest86 from a USB stick overnight. Any red errors mean a faulty stick — and before you shop, read our guide to why RAM is so expensive right now so the replacement price doesn’t blindside you.
Cause 4: drivers and software
If the evidence points software-side — clean restarts, repeatable stop codes like VIDEO_TDR_FAILURE or DRIVER_IRQL_NOT_LESS_OR_EQUAL — start with the graphics driver: uninstall it completely with DDU in Safe Mode, then install the current version fresh. Check Windows Update history against your crash timeline in Reliability Monitor (a driver pushed by Windows Update is a classic trigger), and make sure motherboard chipset drivers are current.
And if you’ve tuned anything — CPU overclock, GPU undervolt, curve optimiser — return it all to stock before diagnosing further. Silicon margins shrink with age and heat, and a PC that keeps restarting under a tune that passed every stress test last year may simply have run out of headroom. Stock-stable first; re-tune later.
Cause 5: the sneaky electrical stuff
The embarrassing-but-real list we check before blaming expensive parts: a worn wall socket or cheap extension strip sagging under load (plug the PC straight into the wall to test), a kettle lead that’s loose in the PSU inlet, GPU power connectors — especially 12VHPWR — not clicked fully home, and the genuinely sneaky one: a faulty case power or reset button shorting intermittently. Unplugging the reset switch header from the motherboard costs nothing and has solved more “haunted PC” cases than we can count.
When to bring it in
Here’s the honest limit of home diagnosis: intermittent faults that survive everything above usually come down to the PSU or the motherboard — and you can’t prove either without swapping in known-good parts. That’s precisely what a bench is for. If your PC keeps restarting after the checks in this guide, or you’d rather not open the case at all, bring it to us in Harrow: we’ll reproduce the fault, isolate the part, and quote before touching anything — no fix, no fee.
Related reading: brand-new build misbehaving instead? Start with our new PC not booting guide. Stutters rather than shutdowns? See the PCIe stutter fix. And if you suspect your parts simply aren’t keeping up, the bottleneck calculator gives a straight answer.
PC still restarting or cutting out? Message us what it’s doing and when — mid-game, at idle, after how long. We diagnose random restarts and shutdowns every week in Harrow.
📱 Get it diagnosed