Your PC is rock solid on the desktop, then reboots with no blue screen the moment a game hits full load. A random restart under load almost always points to one of three culprits: power delivery, heat, or an unstable overclock or memory profile. Here is how to isolate it methodically instead of guessing and swapping parts.
Why load triggers the reboot
Idle draws maybe 60-90 watts. A gaming load can spike a mid-range system to 400-500 watts in milliseconds. A weak, failing, or undersized power supply cannot hold voltage during those transients, so the board trips and restarts to protect itself. Heat and unstable RAM timings fail the same way, only under stress.
Start with the most likely cause
An instant reboot with no error usually means power or a protection trip, not software. A reboot that follows a brief freeze or artifacts leans toward GPU or thermals.
| Symptom pattern | Likely cause | First test |
|---|---|---|
| Instant reboot, no warning | PSU transient trip | Try a known-good higher-watt PSU |
| Reboot after temps climb | CPU or GPU overheating | Log temps; watch for 90 C+ |
| Reboot only after enabling XMP/EXPO | Unstable memory profile | Disable profile, retest |
| Reboot with artifacts first | GPU or its power cable | Reseat 12VHPWR/8-pin, drop GPU clocks |
The diagnostic sequence
Check temperatures under load
Run a monitoring tool and play until it reboots. A CPU or GPU hitting 95-100 C right before the restart confirms thermals: reseat the cooler, improve case airflow, or repaste. Most parts throttle before they cut power, but a marginal cooler mount can spike fast enough to trip.
Rule out the power supply
If temps look fine, suspect the PSU, especially a unit under 650 watts paired with a modern GPU, or any old or bargain unit. A quality 650-750 watt 80 Plus Gold supply gives modern mid-range builds the transient headroom they need. Borrowing a known-good PSU is the fastest confirmation.
Test memory stability
If the reboots started when you enabled XMP or EXPO, turn it off and retest. If stable, the kit or memory controller is not happy at the rated speed; try the next speed tier down or run a memory test overnight.
Isolating the fault without buying parts
Before spending money on a new PSU or cooler, you can narrow the cause with free tests that change one variable at a time. The goal is to make the reboot happen on demand or stop it entirely, which tells you which subsystem is responsible.
Cap the power draw to test the PSU theory
In your GPU software, set a frame rate limit or a modest power limit of around 80 percent and replay the scene that caused reboots. If a lower power ceiling stops the crashes, the power supply cannot handle the transients and needs replacing with a stronger, higher-quality unit. If reboots continue regardless, power is likely not the cause.
Separate CPU load from GPU load
Run a CPU-only stress test and a GPU-only stress test individually. A reboot under CPU load points at the cooler, VRM, or EPS power; a reboot only under GPU load points at the card or its cables. If only the combination crashes, it is almost certainly total-system power delivery, reinforcing the PSU diagnosis.
Check the wall side too
Not every reboot originates inside the case. A failing wall outlet, an overloaded power strip, or brownouts can drop voltage enough to reset the PC. Plugging into a different circuit, or better a UPS that reports input voltage, rules out household power before you blame the build.
FAQ
Could it be a software or driver problem?
A pure instant reboot is rarely software; drivers usually cause crashes to desktop or blue screens, not clean power cycles. Still, update the GPU driver and chipset drivers, and check that Windows fast startup is not masking a shutdown issue before opening the case.
How do I know if my PSU is too weak?
Add up your CPU and GPU rated power and target a PSU with roughly 30-40 percent headroom above that total. A build pulling around 450 watts peak is comfortable on a 650-watt unit; running it on a tired 500-watt supply invites exactly these transient trips.
Bottom line
Log temperatures first, then suspect the power supply, then test the memory profile. Random reboots under load are a process of elimination, and in most builds a quality 650-750W PSU plus a properly seated cooler ends the problem.
