Power · Guide

Can software tell if your PSU is dying?

It can reveal a pattern. It cannot certify a power supply. Those are two very different things.

By Marcin Firmuga·Published 2026-07-29·9 min read
Read this first. If you smell burning, see smoke or sparks, hear repeated clicking, or the machine shuts off and will not power back on, stop. Shut it down at the wall and do not open the power supply. PSU capacitors can hold a dangerous charge long after it is unplugged. This is the one part inside a PC you do not open.

Somewhere in a monitoring app, your +12V rail reads 11.86V, and your stomach drops. Before you add a new power supply to your cart, read this. That number is probably fine, and even if something is wrong, the app on your screen cannot prove it.

A motherboard voltage sensor is not a lab instrument. It can be noisy, it can be mislabelled, and it is routinely off by a few percent. Software can reveal a pattern over time. It cannot certify a power supply. Confusing those two is how people replace a perfectly healthy PSU, or keep running a dying one because a single reading looked fine. This guide is how to read the rail like an engineer, then how PC Workman does the same on your own machine's history.

In this guide
  1. What a "normal" 12V reading actually is
  2. Why one low number is not a verdict
  3. The real signs a power supply is failing
  4. How to actually confirm it
  5. How PC Workman reads the trend
  6. What software can and cannot claim

What a "normal" 12V reading actually is

The ATX specification allows each rail a tolerance of plus or minus 5 percent. For the +12V rail that means anything from 11.40V to 12.60V is in spec. So 11.8V, the number that made your stomach drop, is comfortably inside the allowed range.

Two more facts change how you should read that number:

Why one low number is not a verdict

Modern hardware draws power in sharp bursts. When your GPU spikes, the rail can dip for a fraction of a second and recover. The newer ATX 3.0 spec even permits brief +12V dips to around 11.16V during those transient spikes. A momentary flicker on a graph is not a dying PSU.

The difference that matters is transient versus sustained. A dip that snaps back is the supply doing its job under a sudden load. A rail that settles lower and lower over weeks, or sags and stays sagged whenever the machine works hard, is the pattern worth acting on. One reading is a photo. You need the film.

The real signs a power supply is failing

Voltage on a screen is the least reliable of these. The symptoms that actually point at a PSU are behavioural:

Notice that a single low idle reading is not on this list. The rail behaving badly under load, over time, is what tells the story, which is exactly what a health check that watches trends is built to catch.

How to actually confirm it

If the behaviour above shows up, software has done its job by pointing you at the PSU. Confirming it is hardware work:

  1. Measure it directly. A multimeter on the PSU's connectors while the system is under load gives a far more accurate reading than any sensor. If the rail holds near 12V under load on a meter, the software number was noise.
  2. Cross-check the source. Compare the motherboard reading against a second sensor or tool. Two independent readings agreeing is worth more than one alarming one.
  3. For a real verdict, load-test it. A proper PSU test uses adjustable loads and an oscilloscope to check ripple and transient response. That is the lab-grade answer software can never give.

How PC Workman reads the trend

A single voltage number is close to useless. A pattern in your own machine's history is where the real signal lives, and that is what PC Workman watches. Instead of a fixed threshold, it applies statistical process control to your voltage rails: the median and MAD (median absolute deviation), a modified z-score, and Nelson rules, the same industrial method used to catch a process drifting out of control. It is looking for a trend, not a magic line:

Voltage Trend, your +12V rail30-day SPC
In-spec range11.40 to 12.60 V (ATX plus/minus 5%)
Current median11.92 V, within range
30 days ago12.08 V
Trendmedian down 0.16 V in 34 days
Varianceup 2.8x
Abnormal drops7, during load transitions
Unusual trend detected. This is not a PSU diagnosis. Confirm with a second sensor source or a multimeter under load.

Every value stays inside the ATX range, so a threshold-based tool would say nothing. The trend, a median quietly falling and variance climbing, is what a good monitor surfaces months before a rail actually fails.

What software can and cannot claim

Software reveals a pattern. It does not certify a PSU. PC Workman is deliberate about the wording. It reports "unusual trend detected", not "your power supply is failing", because a motherboard sensor cannot prove that and pretending otherwise would send people to buy parts they do not need. It gives you an early, honest signal and tells you exactly how to confirm it in hardware. An alarmist tool that cries "PSU FAILING" over one noisy reading gets muted in a day, and then it protects nothing.

So the next time a rail reads 11.8V, do not panic and do not order a replacement. Check the spec (you are in it), watch whether it is stable or sagging under load over days, and look for the real symptoms, the shutdowns and the Event 41s. Software points. Hardware proves.

Watch the rail, not a single number. PC Workman records your voltage rails over time and flags a genuine trend with statistical process control, median, MAD and Nelson rules, so real drift shows up early without crying wolf over one noisy reading. It always tells you to confirm in hardware. Free, open source, 100% offline. Download for Windows.

Next question: the machine restarts under load, and you are not sure if it is heat or power. Browse the rest of the guides →

MF

Marcin Firmuga

Developer of PC Workman · HCK_Labs

18 months of building a system monitor in public. The voltage SPC behind this guide is the same engine that ships in the app.