Signs your power supply unit is failing and needs replacement
A client brought me a machine last winter with a story I've heard maybe forty times. His PC would shut off mid-game, restart fine, then die again twenty minutes later. He'd already replaced the RAM, reinstalled Windows twice, and was about to buy a new graphics card. The actual culprit was a five-year-old 600W unit that couldn't hold its 12V rail under load anymore. Total cost of the fix: about $70.
Power supply failures are the most misdiagnosed problem in PC repair, and I say that as someone who misdiagnosed plenty of them early on. The PSU sits at the center of everything, so when it degrades, it imitates a dozen other faults. Here's how to tell what you're actually dealing with.
Key Takeaways
- Random shutdowns under load — not at idle — are the single most reliable sign of a dying PSU.
- Software tools (HWiNFO, OCCT) can show you 12V, 5V, and 3.3V rail readings, but motherboard sensors are approximate. A multimeter or dedicated tester gives you the truth.
- ATX spec allows ±5% tolerance on the 12V rail. Anything sagging toward 11.4V under load is a problem, not a quirk.
- Most decent units last 5 to 10 years. Cheap no-name units often fail in 2 or 3.
- Bulging capacitors, a fan that stopped spinning, or a burnt smell mean replace now, not later.
- A failing PSU can absolutely damage your motherboard, drives, and GPU. The protection circuitry is the whole point of buying a good one.
How to tell if your power supply needs replacing
The short answer: if your PC fails specifically when it draws more power, and passes every other test, you're looking at the PSU. That "specifically when it draws more power" part is what separates a real diagnosis from guessing.
The load test that settles it
Run something demanding — a game, a render, a stress test — and watch what happens. A healthy machine gets louder and hotter. A machine with a tired PSU shuts down, freezes, or reboots. I use OCCT's power test for this because it loads CPU and GPU simultaneously, which is exactly the scenario that kills marginal units. Ten minutes is usually enough. If it survives twenty, your PSU probably isn't the problem.
Then check the rails. HWiNFO64 reads the voltage sensors your motherboard exposes. Look at the 12V line under load:
- 11.8V to 12.2V — healthy
- 11.6V to 11.8V — aging, keep an eye on it
- Below 11.4V — failing, replace it
- Wild swings between readings — worse than a consistently low number
One caveat I learned the hard way: motherboard voltage sensors are notoriously imprecise. I once panicked over a 11.5V reading that turned out to be a sensor calibration issue. If you want certainty, a $25 PSU tester or a cheap multimeter on a Molex connector gives you real numbers.
How to tell if the power supply is bad or the motherboard
This is the question that sends people down expensive rabbit holes. The differential is simpler than it looks.
| Symptom | Likely PSU | Likely motherboard |
|---|---|---|
| Shuts down under heavy load | Yes | Rarely |
| Fails to POST at all, no fans, no lights | Possible | Possible |
| Random reboots at idle | Occasionally | More likely |
| USB ports dropping out intermittently | Yes (5V rail) | Yes |
| Burning smell near the top rear of the case | Almost certainly | No |
| PC works fine after a cold start, dies when warm | Yes | Sometimes |
Here's the trick I use: borrow a known-good PSU and swap it in. Twenty minutes of work, and it eliminates all ambiguity. If the symptoms vanish, you had your answer. If they persist, you saved yourself the cost of a part you didn't need. That single step has saved me more wasted money than any diagnostic tool.
What are the symptoms of a failing power supply?
Symptoms come in two flavors: the ones you notice, and the ones you have to look for. The second category matters more, because a PSU rarely dies dramatically — it degrades quietly for months first.
Visible symptoms
Pull the side panel off and look. Genuinely look, with a flashlight.
- Bulging or leaking capacitors. The tops should be flat. Domed, cracked, or weeping brown residue means it's done.
- A fan that doesn't spin, or spins sluggishly. Some units only spin the fan under load by design — check your model before panicking. But a fan that never moves under load is dead.
- Discolored or melted connectors. Especially the 24-pin and PCIe cables. Heat marks around a connector mean resistance has built up, which means the connection is failing.
- Dust so thick it forms felt. I've opened units where the intake was completely sealed. That alone can cook a PSU in a year.
Behavioral symptoms
These are the ones people describe to me over the phone, and they follow a pattern.
- The machine shuts down under load, then boots normally seconds later
- Coil whine that appeared recently and gets louder under stress
- Intermittent BSODs with no consistent stop code
- Drives disappearing from the OS, then reappearing after a reboot
- The PC won't power on until you flip the switch on the back off and on again
That last one is a personal favorite. I spent two weeks convinced I had a faulty front-panel button. It was the PSU's internal protection latching after a brownout, and cycling the rear switch reset it. Replaced the unit, never saw it again.
What is the average lifespan of a PSU?
Honest answer: it depends almost entirely on quality and heat. A well-built 80 PLUS Gold unit running cool will comfortably hit 7 to 10 years. I've pulled working units out of machines from 2014. Cheap unbranded supplies, the kind bundled into prebuilt office PCs, often start misbehaving at the 2 to 3 year mark.
Heat is the variable nobody respects. Every 10°C above comfortable operating temperature roughly halves capacitor life. A PSU mounted at the bottom of a case with decent airflow will outlive an identical model suffocating under a carpet of dust. If your case has a filter, clean it. That single habit is the cheapest PSU insurance that exists.
What causes a power supply to fail early?
From what I've seen repairing machines over the years, the causes rank roughly like this:
- Poor ventilation and dust accumulation — the overwhelming majority
- Cheap components, especially low-grade capacitors
- Running near or above rated capacity for extended periods
- Dirty mains power, brownouts, and no surge protection
- Age, plain and simple
That third one deserves attention. If you built a system with a 550W unit and later dropped in a GPU that pulls 300W on its own, you're now asking the PSU to run at 90% capacity continuously. It'll do it — for a while. Then it won't.
How common is it for a PSU to fail?
More common than the internet admits, less common than the internet fears. When I look at machines I've serviced where a hardware fault was the cause, the PSU accounts for roughly a quarter of them. That's a rough figure from my own records, not a survey — but it lines up with what other techs report. It's one of the three most frequent failure points alongside storage drives and graphics cards.
The important distinction is how they fail. Most don't die cleanly with a pop and a smell. They drift, providing unstable voltages that cause mysterious crashes for months before you connect the dots. That slow decline is why PSU problems get blamed on Windows, on drivers, on RAM, on everything except the actual culprit.
Can a bad power supply damage your computer?
Yes, and this is the reason I don't tell people to "just run it until it dies."
A degrading PSU can push out-of-spec voltage into your motherboard, your drives, and your GPU. When it fails catastrophically, cheap units without proper protection can take the whole system with them. I've seen a single dead PSU kill a motherboard and two SSDs in the same event. The replacement bill was over $400 for a part that would have cost $80 to swap preemptively.
Good units with over-voltage, over-current, and short-circuit protection usually fail safely — they shut down instead of cooking your components. That protection is most of what you're paying for when you buy from a reputable brand. It's not marketing.
When to replace instead of wait and see
My rule, after too many avoidable disasters: replace if the unit is over 6 years old and showing any symptom, if you can see damaged capacitors, if a connector has heat marks, or if your 12V rail sags below 11.4V under load. Also replace immediately if you've recently added hardware that pushes you past 80% of the unit's rated wattage.
For everything else — an occasional odd reading, a unit that's three years old and behaving — measure, monitor, and wait. Panic-replacing good hardware is its own kind of waste.
The uncomfortable truth is that the PSU is the one component you'll never think about until it ruins your week. It sits there, unseen, converting your wall power into something your expensive parts can use, and it gets zero credit for years of quiet service. Treat it like the foundation it is. When it starts sending signals, listen — because the alternative is finding out the hard way, and that lesson costs a lot more than a power supply.