A machine that will not start: no power, no POST, no display

Why this matters

"It will not turn on" is the most common hardware call there is, and it is three completely different faults wearing one description. A machine with no power, a machine that powers on but does not complete its self-test, and a machine that is running perfectly with nothing on the screen all produce the same sentence from a user.

Separating them is the first move, it costs nothing, and it eliminates two-thirds of the possibilities immediately. Technicians who do not separate them start swapping parts, which is slow, expensive and sometimes destroys evidence about what actually failed.

This lesson applies lesson 34's method to the first and most fundamental class of hardware fault.

The lesson

Separating no power from no POST from no display, which are three different faults

Ask the machine three questions in order.

1. Is there any sign of power at all? Fans, lights, drive noise, a fan that spins for a second and stops.

• Nothing whatsoever → a power fault. Mains, cable, supply, switch, or the front-panel connector from lesson 27.
• Power present → move on.

2. Does the machine complete its power-on self-test? Listen and look for beeps, diagnostic lights, a display of the manufacturer's logo, drive activity after the initial spin-up.

• Power but no POST → the machine is not getting far enough to use a display. Processor, memory, board, or firmware.
• POST completes → move on.

3. Is there an image?

• No image but the machine clearly boots — drive activity settles, network lights behave normally, the machine responds to a keyboard shortcut that would shut it down → a display fault, which is lesson 38's territory and the cheapest of the three to fix.

Making this split takes under a minute and it is the difference between a directed diagnosis and a shopping trip.

Power-on self-test indicators: beeps, codes, and boards that show neither

The self-test checks the essential components before handing over to an operating system, and it reports failures in whatever way the board offers.

• Beep codes. A speaker — either on the board or a small one connected to the front-panel header — signals a pattern. One short beep usually means all is well; repeating long beeps often mean memory; patterns vary by firmware vendor, so the board's manual is the reference. Many modern desktops have no speaker fitted, which is why the code that would have told you the answer is silent. A cheap board speaker is worth having in a test kit.
• Diagnostic LEDs. Most current boards have four labelled lights — CPU, DRAM, VGA, BOOT — that light in sequence and stop at whatever failed. This is the fastest diagnosis available on a modern machine and requires only that you look at the board while it starts.
• POST code displays. A two-digit hexadecimal readout, common on enthusiast and server boards, giving a precise stage.
• Laptops and branded desktops often use keyboard lights or a coded blink pattern on a status LED, documented by the manufacturer.

So the first action after establishing "power but no POST" is not to open anything. It is to look for what the machine is already telling you.

The minimum configuration test, and what it proves when the machine starts

When there are no indicators, or they point nowhere, reduce the machine to the smallest set of parts that can complete a self-test:

• Motherboard, processor and cooler.
• One memory module, in the slot the manual names as first.
• Graphics — the integrated one if the processor has it, otherwise one card.
• Power supply.
• Nothing else. No drives, no expansion cards, no front-panel USB, no extra fans, no peripherals.

Then start it, ideally outside the case on a non-conductive surface if a short against a standoff is suspected.

• It starts → the fault is in something you removed. Add components back one at a time, starting each time, until it fails. The component that reintroduces the fault is the fault.
• It still does not start → the fault is in what remains: board, processor, memory module, graphics, or supply. Swap the memory for another module and another slot, then substitute the supply, then the graphics.

This test is powerful because it turns an unbounded problem into a bounded one. A machine with fourteen components has too many combinations; a machine with five has very few.

Reseating as a diagnostic step rather than a ritual

Reseating works often enough that people do it reflexively, which obscures why it works. Connections fail because of vibration, thermal cycling, oxidation on contacts, transport, and installation that was never quite right.

Reseat, in this order:

1. Memory. The most common beneficiary by a distance, and the classic symptom is a machine that was moved and now does not POST.
2. Graphics card, including its power connectors.
3. Power connectors to the board, especially the CPU power connector that lesson 28 flagged as the one people forget.
4. Drive data and power cables.
5. Front-panel header, particularly if the machine shows no response at all to the power button.

Doing it as diagnosis rather than ritual means changing one thing and retesting. Reseating everything and finding it works tells you nothing about what was wrong, and the fault will return without anyone knowing where to look.

On the mechanical side, look while you are in there: bulging or leaking capacitors on the board, burn marks, a standoff under the board where there should be none, a cooler that is loose, and dust packed into the heatsink.

Deciding between a board, a supply and a processor without a spare of each

This is the hard case: a machine that will not POST in minimum configuration, with no diagnostic indicators, and no spare parts.

What can be established without substitution:

• Does the supply respond? With the machine off but plugged in, the board's standby light should be lit on boards that have one. Its absence points at the supply or the mains path.
• Does the supply start? Fans spinning means it is producing something, though not necessarily correctly. Lesson 43 covers measuring the rails, which is where a multimeter turns this from an opinion into a reading.
• Is the processor getting hot? A processor being powered warms slightly within a minute or two. Stone cold after several minutes of apparent power suggests it is not being powered at all, which points at the board or the CPU power connector.
• Does removing all memory change the behaviour? A working board with no memory should complain — beeps, a memory LED, something. A board that behaves identically with and without memory is not getting far enough to check, which points at the board, the processor or its power.

Probabilities worth knowing, since they order the substitution attempts: power supplies fail most often, memory next, then motherboards, and processors very rarely. A processor is almost never the answer unless it was physically damaged during installation — which is why lesson 18 spends so long on that.

The practical decision: borrow or buy a known-good power supply first, because it is the most likely and the most reusable test item. If the machine is under warranty, stop and use it — and note that on many branded machines opening the case is itself a warranty question, which belongs in step four of the method as an effect of the plan.