A flaky power supply is one of the hardest faults to diagnose because it hides behind other symptoms: random reboots, black screens under load, and crashes that look like GPU or driver problems. If your gaming PC shuts down or restarts specifically when the action heats up, the PSU deserves scrutiny. Here is how to stress-test and evaluate power delivery without specialized lab gear.
Recognize the symptoms of a weak PSU
Power problems tend to appear under transient load, the split-second spikes a GPU draws during demanding scenes. Classic signs include the PC rebooting or powering off entirely when a game gets busy, the display dropping to black while the system stays on, or crashes that only occur in graphically intense titles while lighter games run for hours. A clean software crash is different from an instant power-off; the latter points at power.
| Symptom | PSU suspicion | Cross-check |
|---|---|---|
| Instant power-off under load | High | Temps, wattage headroom |
| Reboot with no BSOD | High | Kernel-Power event 41 |
| Black screen, system stays on | Medium | GPU power cable, driver |
| Crash only in heavy games | Medium | GPU transient spikes |
Check the basics first
Wattage and headroom
Add up your components’ realistic draw and compare it to the PSU rating. A modern midrange build wants a quality 650-750W unit; higher-end cards with big transient spikes are happier on 850W or more. Running a supply near its limit invites shutdowns during load spikes even if the average draw looks fine. Aim to keep sustained load under roughly 80 percent of the rated wattage.
Cabling
Use the PSU’s native PCIe power cables, and for power-hungry cards use separate cables rather than a single daisy-chained one with two connectors. Reseat the 24-pin, the CPU EPS connector, and the GPU cables. A loose or marginal connection produces exactly the intermittent power-off behavior people blame on the whole unit.
Stress-test under controlled load
Run a combined load
Use a GPU stress tool alongside a CPU stress test to push both rails at once, since real games load them together. Monitor for shutdowns or reboots over a 30-60 minute run. If the system fails a combined test but survives each part alone, that strongly implicates power delivery rather than either component individually.
Watch the software rail readings
Hardware monitoring utilities read the 12V, 5V, and 3.3V rails. These sensor values are approximate, but a 12V rail that sags well below 12V (for example under about 11.4V) under load is a warning sign. Combine this with the crash timing from Event Viewer’s Kernel-Power entries to build a case. Consistent deep sags plus power-offs mean the unit cannot hold voltage.
Confirm by swapping
The most conclusive test is substitution. Swap in a known-good, adequately rated PSU and see if the crashes vanish. Power supplies also age; capacitors degrade over years, so a five-plus-year-old budget unit that was borderline when new may no longer hold up to a modern GPU. If a good replacement fixes it, the old unit was the fault.
Account for GPU transient spikes
Average power draw understates what a modern graphics card actually demands. Many cards produce transient spikes lasting only milliseconds that reach roughly double their rated draw, and a protective PSU can trip on these spikes even though the average load looks safe. This is why a card with a 250W rating is happiest on a supply with real headroom and good transient handling rather than one sized to the average. If your PC shuts off at the exact moment a scene explodes with effects, suspect transients. A higher-quality unit or more wattage headroom absorbs these spikes without tripping over-current protection, curing shutdowns that no software setting could.
FAQ
Can software fully test a power supply?
Not completely. Software can log rail voltages and reveal sags, and stress tools can provoke failures, but sensor readings are approximate and cannot measure ripple or true transient response. Software narrows it down; a physical swap with a known-good unit is the definitive test.
How much PSU headroom should a gaming PC have?
Plan for sustained load to stay around 70-80 percent of the rated wattage, leaving room for GPU transient spikes that briefly exceed average draw. For a typical midrange build that means a quality 650-750W unit; high-end cards with large spikes are safer on 850W or more.
Bottom line
Suspect the PSU when the PC powers off or reboots under load without a blue screen. Verify wattage headroom and cabling first, then run a combined CPU and GPU stress test while watching rail voltages and Kernel-Power events. Because software can only go so far, confirm with a known-good replacement, and remember that aging units lose capacity over the years.
