A PC That Will Not POST

Work the list in order of probability, not in order of what you are afraid of. The overwhelming majority of no-POST machines are a connector, a stick of memory, or something shorting — and the parts people suspect first are the ones that fail least.

A note on this guide: this rests on how POST works and on which failures are structurally most likely, not on machines broken deliberately here. See how we test.

The problem, as it appears

You press the button. Fans spin, lights come on, and nothing appears on screen. No beeps, or beeps you cannot interpret. The machine looks alive and is not.

That combination — clearly powered, clearly not starting — is what makes people conclude a major component has died. It is also exactly what a missing connector looks like.

Why "fans are spinning so it has power" disappoints

Because fan spin proves one thing only: the power supply switched on.

The supply comes up when the board tells it to, and that happens early. Everything after it — memory training, the CPU initialising, the board handing over to firmware — can fail completely while the fans continue turning happily.

So spinning fans is not evidence of a working board, a working CPU, or adequate power reaching either. It is evidence of a working power switch.

The list, in order

Work down it. Change one thing at a time and try again, because changing two means learning nothing from the result.

1. The CPU power connector

The separate four or eight pin connector at the top of the board, distinct from the large one down the side. Push it home properly — it is awkward to reach after the board is in, and it seats less convincingly than the main connector.

This is the most missed item on the list by a wide margin, and its symptom is precisely fans-and-no-display.

2. Memory, down to one stick

Remove everything but a single module and put it in the slot the manual specifies for one-stick operation, which is usually not the one nearest the CPU.

Reseat rather than just look. Contacts that have been fine for years can become marginal through thermal cycling, and this test costs nothing.

3. Everything else out

Graphics card, drives, extra fans, front-panel USB. Get to board, CPU, cooler, one stick, and the power supply.

If it posts, add things back one at a time. The component that stops it is the answer, and you will have found it without buying anything.

4. Short the power pins directly

Front-panel headers fail and front-panel buttons fail. Briefly bridging the two power pins on the header with a screwdriver starts the board directly and eliminates the case's wiring from the picture.

If it starts that way and not from the button, the fault is in the case rather than in the machine.

5. Take it out of the case

A standoff in the wrong hole shorts the underside of the board against the tray, and it is close to invisible once assembled.

Build it on the box the board came in and try again. This one test rules out both standoffs and any other contact between board and case.

6. Clear CMOS

A bad memory overclock or a firmware setting that the board cannot recover from will stop a POST, and clearing resets it. Use the jumper or button if the board has one; otherwise remove the coin cell with the machine unplugged.

7. Then, and only then, the expensive parts

Power supply, board, CPU. All three are real causes and all three are less common than everything above.

A tester tells you something about the supply, with an important limitation.

This page contains affiliate links. If you buy through them, Ugly Fixes may earn a commission at no extra cost to you. Products are selected independently.

What actually matters

One change at a time

The temptation under frustration is to reseat everything, swap two parts and try again. If it works you have learned nothing and you do not know what to avoid next time.

Beeps and diagnostic lights are the free information

Many boards have a speaker header or a set of stage LEDs, and both tell you which phase failed — CPU, memory, graphics, boot. That is the difference between working a list and reading an answer.

If your board has a speaker header and no speaker, that is a component worth owning before you need it rather than after.

Fitting things right prevents most of this

A substantial share of no-POST machines are assembly errors rather than failures, and the tools that prevent them are cheap — PC upgrade tools that prevent damage covers the set.

The mistake to avoid

Buying a part to test a theory. It feels like progress and it is the most expensive form of guessing available, because a machine that does not POST gives you no confirmation that the part you replaced was the one at fault. Work the free list first — it costs an hour and it is right most of the time.

The second mistake is concluding the CPU has died. It is close to the least likely thing on the list, it is the most expensive to replace speculatively, and the failure people attribute to it is nearly always the connector that feeds it — which is the first item above for exactly that reason.