A new power supply is the one component you want to catch failing early, ideally within a return window, not six months into ownership when it takes hardware down with it. Stress testing, sometimes called burn-in, deliberately pushes the PSU to reveal instability, poor voltage regulation, or infant-mortality defects before you trust it with daily use. You do not need lab equipment; software and careful monitoring get you most of the way. Here is a safe, practical approach.
Why burn-in matters
Electronic components that are going to fail early often do so under load in their first hours of operation, a pattern called infant mortality. A structured stress test surfaces these weak units while you can still return them, and it confirms your specific unit holds voltage steady under the transients your GPU and CPU generate. A PSU that sails through a hard test is one you can stop thinking about.
What to watch
The goal is to load the system hard while monitoring three things: voltage stability on the 12V rail, system stability (no crashes or reboots), and unusual behavior like coil whine, clicking, or shutdowns.
| Metric | Healthy | Warning sign |
|---|---|---|
| 12V rail under load | 11.4-12.6 V (within 5%) | Dips below ~11.4 V |
| System behavior | Stable, no reboots | Random shutdowns / restarts |
| Noise | Quiet fan hum | Clicking, buzzing, loud whine |
| Smell / heat | Warm exhaust | Burning smell, hot to touch |
Step by step
1. Set up monitoring
Open a hardware monitoring tool and watch the +12V, +5V, and +3.3V readings. Software voltages are approximate but a large sag is still a clear red flag. Note the idle values as a baseline.
2. Load the CPU and GPU together
Run a CPU stress utility and a GPU stress load simultaneously to push the PSU toward its real peak draw. This combined load stresses the 12V rail hardest, which is exactly what you want to test. Run it for 30-60 minutes.
3. Add transient stress
Alternate a demanding game or a benchmark loop that repeatedly loads and unloads the GPU, since transient spikes, not steady load, expose a weak supply. Watch for reboots or voltage dips as the load swings.
4. Monitor and judge
A healthy unit holds 12V within tolerance, never reboots, and stays quiet. Random shutdowns, voltage sagging under 11.4 V, loud clicking, or any burning smell means stop immediately and return the unit.
What software can and cannot tell you
Be honest about the limits of a software-based burn-in. Motherboard voltage sensors are approximate, sometimes off by a tenth of a volt or more, so treat the readings as trend indicators rather than lab-grade measurements. That is why independent reviews with proper test benches remain the gold standard for judging a PSU’s true quality.
Why this test is still worth doing
Even without lab gear, a home stress test reliably catches the failures that matter most to you: a unit that reboots under load, sags badly, clicks, or shuts down. Those are exactly the defects that turn up in the first hours of use, and catching them inside a return window is the whole point. Think of burn-in as screening for obvious faults, not certifying perfection, and it earns its 30-60 minutes many times over.
FAQ
How long should I burn in a new PSU?
A combined CPU and GPU load for 30-60 minutes, plus some transient gaming or benchmark cycling, catches most early failures. Some builders run an overnight looped stress test for extra confidence. The key is sustained heavy load plus load-swing transients, not just idling the system for hours.
Is stress testing a PSU safe?
Yes, on a quality unit, because you are only loading it to what your hardware can actually demand, which is within its rating. The test simply reveals a defective unit sooner. Stop at once if you smell burning, hear persistent clicking, or see repeated shutdowns, as those indicate a genuine fault, not a passed test.
Bottom line
Burning in a new power supply means loading the CPU and GPU together, adding transient swings, and watching the 12V rail and system stability for 30-60 minutes or more. A good unit holds voltage within 5%, never reboots, and stays quiet, while a defective one reveals itself with sags, shutdowns, or noise, ideally while you can still return it. It is a small time investment to protect every other component in the build.
