PSU Failure Symptoms: Is It Really the Power Supply? (2026)

By LK Wood IV · 2026-07-10 · ~9 min read · St. Louis County, MO

Some retailer links below are Amazon affiliate links; if you buy through them TechFuel HQ may earn a small commission at no extra cost to you. As an Amazon Associate I earn from qualifying purchases. Commissions never influence what gets recommended — see our disclosure.

The power supply is the hardest part in a PC to diagnose, because when it starts to fail it doesn’t announce itself — it frames another component. Random shutdowns look like an overheating GPU. Blue screens look like bad RAM. A machine that won’t POST looks like a dead motherboard. People swap out card after card, reinstall Windows, reseat the memory, and burn a weekend — and the culprit was quietly sagging under the whole time.

So before you spend anything, the real first question isn’t “what are the symptoms of a failing PSU” — it’s “is it actually the PSU, or something the PSU is being blamed for?” Get that right and the rest is easy.

The signs of a failing power supply

Here’s what a dying PSU actually does. None of these alone is proof — several of them overlap with other faults — but the pattern is the tell.

• Random shutdowns or hard reboots, usually under load. The PC works fine browsing or on the desktop, then cuts out completely the moment you launch a game, start a render, or otherwise spike the power draw. No blue screen, no error — it just dies, sometimes rebooting on its own. (If that no-warning, no-BSOD shutdown is your main symptom, our computer randomly turns off guide walks the full heat-vs-PSU-vs-short diagnosis in order.)
• It won’t turn on at all, or turns on for a second and stops. Press power and get nothing — no lights, no fans — or a brief twitch that immediately shuts back down. (A dead press has cheaper causes too — a loose front-panel wire or the button itself — so rule those out first with Power Button Not Working on Your PC? before you condemn the PSU.)
• Blue screens and instability with no consistent cause. BSODs that point at different drivers each time, freezes, and USB devices dropping out can all trace back to unstable power delivery.
• Coil whine or buzzing that rises with load, or grinding from the PSU fan as its bearing wears. (Occasional light coil whine is normal and harmless on its own — it’s new or worsening buzz paired with the crashes below that points at the PSU.)
• A burning or electrical smell — the most serious sign. Unplug immediately (more on that below).
• The “random” fingerprint. This is the big one: a failing PSU crashes inconsistently. It dies in a game today, a video export tomorrow, then behaves for days. Hardware that fails on a schedule usually isn’t the PSU; hardware that fails at random, under load, with no error, very often is.

If you recognize the load-dependent, no-error, random pattern, the power supply moves to the top of the suspect list. Now rule the impostors in or out.

Step 1: Is it the PSU, the GPU, or the motherboard?

These three produce nearly identical top-level symptoms — sudden shutdowns and crashes — so the diagnosis is about the secondary clues each one leaves.

The reason PSU faults feel “random” is physics: a degrading unit still holds voltage fine when the system is sipping power at idle, and only drops out of spec when a modern GPU slams it with a sudden current spike. That’s why the same PC can be rock-solid for hours of light use and then die ninety seconds into a game. Power-supply maker Seasonic describes the same split — PSU trouble shows up as sudden shutdowns under load with no software error, while a failing GPU leaves visual or driver clues first.

If the crashes come with visual glitches or driver resets, work the GPU angle instead: GPU Artifacts? Sort Software From a Dying Card. If the PC won’t display or POST at all, start where the boot chain breaks: Graphics Card Not Detected? Find Where It Disappears. And if it’s restarting rather than dying dark, the broader checklist lives in PC Keeps Restarting? Work Through the Causes.

The gold-standard tie-breaker, if you can manage it: swap in a known-good power supply — borrowed, or a cheap-but-reputable spare. If the shutdowns vanish, you’re done, no guesswork. It’s the single most conclusive PSU test there is, which is why experienced builders reach for it before any meter.

Step 2: Why the paperclip test lies

Search “how to test a PSU” and the top answer is almost always the paperclip test (also called the jumper test): unplug everything, leave only the 24-pin cable, and short the green PS_ON pin to any black ground pin with a bent paperclip. Plug in, flip the switch, and if the fan spins, the internet declares your PSU “good.”

It isn’t wrong, exactly — it’s just answering a much smaller question than you think. All the paperclip test proves is that the PSU can switch on. It runs the unit with essentially no load, and a failing supply almost always passes at idle. The way these units die is by failing to hold voltage under load — precisely the condition the paperclip test never creates. A power supply that spins its fan on the bench and then crashes your PC the instant a game loads has “passed” a test that measured the wrong thing.

So treat a passed paperclip test as “not completely dead,” not “healthy.” For a real read you need to load it and watch.

Step 3: Actually measure it — under load

Here’s the testing ladder from easiest to most conclusive.

1. Watch the 12V rail under load (free, best first move). Install HWiNFO (free), open Sensors, find the 12V, 5V, and 3.3V readings, and run a stress test — OCCT has a dedicated power-supply test, or run a demanding game. The ATX spec allows each rail a ±5% tolerance, so the valid windows are:

Watch the 12V rail especially — it feeds the CPU and GPU and is where trouble shows first. Steady readings inside the window are healthy; a 12V line that sags toward or below 11.4V when load hits, or swings around erratically, is a failing supply caught in the act. (Motherboard sensors aren’t lab-accurate, so read the trend and stability more than the third decimal — the newer ATX 3.0 spec even permits brief 12V dips to about 11.16V during transient spikes, so a momentary flicker isn’t damning; a sustained sag is.)

2. Use a PSU tester (~$20–30). A dedicated PSU tester plugs onto the PSU’s connectors and displays each rail’s voltage. It’s more convenient than a multimeter and catches a rail that’s dead or badly out of spec. Its limit is the same as the paperclip test’s — it reads at low load — so a clean result doesn’t fully clear a unit that only misbehaves under a real GPU spike. Good for catching gross faults, not subtle sag.

3. Multimeter on the rails. If you own a multimeter, you can back-probe the 24-pin against the ranges above. Same caveat: it’s a snapshot, and low-load.

4. The swap test (most conclusive). As in Step 1 — a known-good PSU in place of the suspect one settles it completely. Nothing beats it.

Notice the theme: every bench test is limited because it can’t recreate a real load, which is exactly when a dying PSU shows itself. That’s why watching the rails under an actual stress load and swapping in a known-good unit are the two tests that actually earn their conclusions.

Safety: the one part you never open

A power supply stores lethal energy. The large capacitors inside can hold a dangerous charge for a long time even after it’s unplugged — enough to seriously hurt you. There is nothing user-serviceable in there, so never open the PSU casing. All the testing above happens from outside the unit, at the cables and connectors. If a PSU is bad, you replace it; you don’t repair it.

And if you ever get a burning smell, smoke, a loud pop, or see sparks — power off at the switch and pull the plug from the wall right away, and don’t turn it back on. At that point diagnosis is over; the unit is done.

When to stop testing and just replace it

Power supplies are the one component where “just replace it” is often the smart move rather than a cop-out, for three reasons:

• A quality unit is affordable relative to what it protects — a good 650–850W supply is a fraction of the cost of the GPU and board it feeds.
• A failing PSU can take other parts with it. An out-of-spec unit can corrupt data or, in a real failure, push a spike into your motherboard, GPU, or drives. Continuing to run a PC you suspect is being fed bad power is a gamble with expensive stakes.
• Diagnosis has real limits. Short of the swap test, you can’t fully prove a PSU is healthy — only that it isn’t obviously dead. If the symptoms fit (random shutdowns under load, no error, an older or cheap unit) and you can’t borrow a spare, buying a reputable replacement is often cheaper than the hours you’d spend chasing ghosts.

When you do replace, buy quality and leave headroom: a reputable unit with a solid warranty, an 80 PLUS efficiency rating, and enough wattage that you’re not running near the ceiling. If you’re not sure how much you need, size it to your actual parts plus a comfortable margin with the PSU calculator rather than guessing high or low. And for which specific unit to buy at that wattage, our PSU tier list ranks 28 current supplies with every placement sourced to a professional test. A good PSU is the least glamorous upgrade in the machine and the one most worth not cheaping out on — it’s the part every other part depends on.

The quick version

1. Recognize the fingerprint. Random shutdowns or reboots under load, with no error message, that come and go unpredictably — that’s the classic failing-PSU pattern.
2. Rule out the impostors first. Visual glitches or driver resets → suspect the GPU. Won’t POST → start at the boot chain. Idle shutdowns that survive a known-good PSU → possibly the motherboard. Memory-named BSODs → test the RAM.
3. Don’t trust the paperclip test. It only proves the PSU turns on, not that it holds voltage under load — which is how these units actually fail.
4. Measure under load. Watch the 12V rail in HWiNFO during a stress test (valid range 11.40–12.60V); a sag under load is the smoking gun. A PSU tester or multimeter catches gross faults; a known-good-PSU swap is the only conclusive test.
5. Never open a PSU — the capacitors hold a lethal charge. Burning smell or a pop → unplug from the wall immediately.
6. When it fits, just replace it. A quality PSU is cheap insurance against a failing one damaging everything it powers.

Sort out whether it’s actually the power supply before you swap a single other part, and you’ll skip the weekend of replacing healthy components one at a time — the exact trap a failing PSU is built to spring.