GPU Takes Several Presses to Power On? Intermittent Boot Failure — PSU, Motherboard & GPU Short Diagnosis Guide

| GPU Repair | HK GPU Fix

GPU Takes Several Presses to Power On? Intermittent Boot Failure — PSU, Motherboard & GPU Short Diagnosis Guide

Key Takeaways

Pressing the power button several times before the PC starts means one link in the power-on chain is in a borderline state, not that a part is dead. This guide explains the start-up sequence, what the fan-spin-then-stop symptom means, and how the remove-the-card comparison separates a graphics card fault from a power supply or motherboard problem.

What This Guide Is — and What It Is Not

This guide is not a general guide to a PC that will not boot. It answers one narrower question: when the PC needs several presses of the power button before it starts, or only starts intermittently, how do you tell whether the graphics card is the cause?

The power supply, the motherboard and the power button are covered here only up to the point where they can be ruled out as the cause. They are not covered as a general repair manual. If the symptom turns out to have nothing to do with the graphics card, the guide says so and stops there, rather than expanding into a full power supply or motherboard repair procedure.

That boundary matters because the symptom itself is not specific to graphics cards. A power supply with an ageing standby circuit, a motherboard with a faulty power sequencing circuit, a worn power button and a graphics card with a power short can all produce the same outward behaviour: press the button, nothing happens, press it again, and eventually the machine starts. The work of this guide is to separate those causes with observation and one decisive comparison, not to guess which part to replace first.

The Power-On Sequence: What Happens After You Press the Button

Before diagnosing anything, it helps to know the order of events that a press of the power button is supposed to trigger. The sequence below is the general order for a desktop PC, not a procedure specific to any brand.

  1. The power supply's standby circuit is live whenever the mains cable is connected. It provides a small standby supply, +5VSB, so the motherboard sits in a waiting state even while the machine is off.
  2. Pressing the power button sends a start signal to the motherboard. The motherboard then pulls the power supply's PS_ON line low, telling the supply to switch on its main outputs.
  3. Once every voltage rail is stable, the power supply sends a Power Good signal. The motherboard then releases the CPU to begin executing, the power-on self-test (POST) runs, the graphics card is initialised, and a picture appears.
  4. If any step times out, is unstable, or trips a protection circuit, the start stops at that step.

The defining feature of the "several presses" symptom is that it sometimes stops and sometimes gets through. That means the failing link is in a borderline state, not completely dead. A part that had failed outright would fail the same way every time. An intermittent failure points to a circuit that is slow to respond, a protection latch that has not reset, or a contact that only sometimes makes a connection.

Step 1: Watch and Record What Each Press Does

The first step is observation, not replacement. Press the button and note exactly what happens. Do this several times, and write down the result of each attempt, because the reaction is not the same every time — and the difference between reactions points to different links in the chain.

Reaction to pressing the buttonWhat it points toNote
No reaction at all — no lights, no fan movement, no soundThe front of the power chain: wall power, power strip, power supply standby circuit, power button, or motherboard standby circuitThe machine has not entered POST; this is not a display output problem
Fans spin briefly (or a light flashes once) and then everything stops, with no response to further pressesThe power supply protection has tripped, or a short circuit on the load side has dragged the supply downOne of the core branches of this guide — see Step 2
Lights on, fans spinning, but no pictureThe power stage has passed; the machine is in the POST or display output stageThis is not the main subject of this guide; it belongs to the guides on a PC that powers on but shows no display
Several presses finally start the machine, and it then runs normallyA borderline state: an ageing power supply standby circuit, a protection latch that has not reset, or a contact problemSee Step 3

Two details are worth recording alongside the reaction. First, whether the machine is harder to start after sitting switched off for a long time than it is when it has recently been running. Second, whether pressing the button immediately after a failed attempt does nothing, while unplugging the mains cable, waiting a while and pressing again makes it start. Both patterns are diagnostic, and both are explained in Step 3.

Step 2: The Decisive Comparison — Booting Without the Graphics Card

The single most useful test in this guide is the remove-the-card comparison. It does not prove the card is dead. It only shows which side the suspicion falls on — and that distinction is the whole point.

Before touching anything inside the case, switch the machine off, unplug the mains power cable from the wall, and wait a while for the residual charge inside the machine to dissipate. Never insert or remove a power cable or a graphics card while the machine is powered on. Every hands-inside-the-case step in this guide starts from that state.

Remove the graphics card from its slot. Then try to start the machine:

Read the result like this:

Do not skip this comparison and go straight to replacing parts. Without it, a power supply fault and a card fault look identical from the outside, and replacing the wrong one changes nothing.

Step 3: The Power Supply Branch

If the remove-the-card comparison points away from the card — or if the pattern of failures already suggests the supply — there are three distinct power supply behaviours to separate.

Ageing standby circuit. In a power supply that has been in service for a long time, the standby circuit can become slow to respond. The first press of the button then misses its timing window: the Power Good signal arrives too late, and the start is abandoned. Pressing again, or unplugging the mains and trying again after a wait, can get through on a later attempt. The characteristic pattern is that the machine is hardest to start after sitting unused for a long time, and once it is running it behaves normally. This is a standby circuit problem, not a question of how powerful the supply is.

Protection latch that has not reset. When a power supply trips its over-current or short-circuit protection, the protection latches on. It stays latched until the supply is fully disconnected from the mains — the mains cable unplugged and the residual charge allowed to dissipate. Pressing the button again and again while the supply is still connected to the wall does nothing, because the latch is still holding. The characteristic pattern is that pressing the button immediately after a failure produces no reaction at all, while unplugging, waiting and pressing again starts the machine. Recognising this pattern matters because it looks like a dead machine, when in fact it is a latched protection circuit waiting for a full power disconnect.

Power cable contact. On a modular power supply, a cable that is not fully seated, or a power connector that has worked loose, can interrupt the power chain intermittently. The mechanism is the same loose-interface mechanism described in the guide on power connector contact problems, borrowed here in one sentence: a connector that is not fully seated can behave normally most of the time and fail on a cold start. Reseating the cables firmly, with the mains unplugged, is the check.

One warning belongs in this section. If the fault is a short circuit on the graphics card dragging the protection down, fitting a larger power supply does not fix it. A larger supply only feeds more current into the same short, which does not solve the problem and carries a risk of burning something. The remove-the-card comparison in Step 2 exists precisely to catch that case before money is spent on a replacement supply.

Step 4: The Motherboard and Power Button Branch

If the card has been ruled out and the supply branch does not fit, the remaining suspects are the motherboard and the power button itself.

The power button. A mechanical button wears with use, and its contacts can become unreliable. The front-panel cable connecting the button to the motherboard can also work loose. If the motherboard has its own power button, that button can be used as a comparison — pressing the board's own button instead of the case button. If the board button works reliably and the case button does not, the fault is in the case button or its cable. What this guide does not teach is shorting the motherboard's power pins with a screwdriver. That is not a diagnostic step for a home user, and it is not described here.

The CMOS battery. A flat CMOS battery is commonly blamed for a machine that will not start, but that blame is overstated. A flat battery normally only causes BIOS settings and the clock to be lost; the machine generally still powers on. The battery should not be presented as the main cause of a power-on failure. At most it is a side check: if the clock resets every time the mains is unplugged, the battery is worth replacing, but it does not explain a machine that needs several presses to start.

Motherboard standby and sequencing circuits. A fault in the motherboard's own standby circuit or in its power sequencing is beyond the scope of home diagnosis. This guide states that limit plainly: once the card, the supply and the button have been checked, a motherboard-level fault is a matter for inspection, not for further home testing.

Step 5: The Graphics Card Branch

When the remove-the-card comparison points to the card, there are two card-side mechanisms to understand.

A power short on the card trips the supply protection. A shorted VRM MOSFET on the card, for example, is a load-side short that makes the power supply trip its short-circuit protection the moment it tries to start — which is why the outward symptom is "fans spin once and then stop". The mechanism is borrowed in one sentence from the guide on VRM MOSFET shorts; this guide does not repeat that guide's material. The important point for diagnosis is that the supply is doing its job: it is protecting itself and the rest of the machine from a short. The supply tripping is evidence of a short somewhere on the load side, not evidence that the supply itself is faulty.

A wrong or shorted power cable. A power cable plugged into the wrong socket — the mechanism described in the guides on modular cable mix-ups and on EPS versus PCIe connectors, borrowed here in one sentence without repeating the pinout detail — can also trigger protection at start-up. The same applies to a damaged cable or a short inside the card's own power connector. The check is to inspect the cable routing and seating with the mains unplugged, and to confirm that each cable is in the socket it belongs in, following the markings on the supply and the card.

One boundary to keep in mind: a card that is completely dead while the rest of the machine powers on normally is a different symptom. That belongs to the guides on a PC that powers on with no display, not to this guide.

SymptomWhere the fault most likely sitsFirst test
No reaction at all to the buttonWall power, power strip, standby circuit, power button, motherboard standby circuitCheck the mains connection, then try the motherboard's own power button if it has one
Fans spin once and then stopPower supply protection tripped, or a short on the load sideRemove-the-card comparison
Starts reliably with the card removedGraphics card side — power short or connector shortInspection to establish the cause of the short
Still needs several presses with the card removedPower supply, motherboard, or power button — not the cardFollow the power supply branch, then the button branch
Hard to start after sitting unused, normal once runningAgeing standby circuitNote the pattern; the supply is the suspect, not the wattage
No response right after a failure, starts after unplugging and waitingProtection latch not yet resetUnplug the mains, wait for residual charge to dissipate, then press again
Clock resets when the mains is unpluggedCMOS battery, as a side check onlyReplace the battery; do not treat it as the main cause

Edge Cases

Cold starts are normal, but warm restarts fail. That is the opposite direction to this guide. The guide on cold-start versus warm-restart behaviour covers it, and this guide borrows only the cold-versus-warm comparison in one sentence. Here the problem is the cold start itself — the machine needs several presses from cold.

The machine starts, then restarts or goes black while gaming. That is a load-stage problem, not a start-up-stage problem. It belongs to the guide on restarts and black screens under load, which borrows only the power supply protection mechanism in one sentence.

A "please connect the PCIe power cable" warning appears on start-up. That is a different symptom: there is a warning message on screen and the machine stops before entering the system. This guide's symptom has no warning message and is intermittent. The warning-message guide is borrowed here in one sentence only, as a contrast.

The machine powers on but stops at the motherboard logo. That is the POST stage, not the power stage. It belongs to the guide on a machine stuck at the logo screen, borrowed here in one sentence as a stage boundary.

Fast Startup affects shutdown and wake behaviour, not the first press of the power button. The fast-start setting changes how the machine shuts down and wakes; it does not change the PS_ON sequence on a cold press. The two should not be confused.

The problem only appeared after changing the CPU or RAM. A part change is a trigger event, and the first thing to check is whatever was touched during that change. The guides on a GPU that is not detected after a CPU change and after a RAM change apply the same logic: a changed part can reset BIOS settings or disturb a connector, and the diagnosis starts by restoring the state from before the change and by checking what was moved. The remove-the-card comparison still applies, and it still comes before any part is declared faulty.

Common Misconceptions

"The PC will not start, so the power supply is not powerful enough." Wrong. Whether a supply is powerful enough and whether the machine starts are two different questions. A supply labelled 750W with a faulty standby circuit can still fail to start the machine, and a machine that fails to start does not prove the wattage is insufficient. No generic wattage threshold is given in this guide as a standard, because wattage is not the variable that decides whether the machine starts.

"Pressing the button several times and eventually getting in means it must be a loose contact." Not necessarily. A protection latch that has not reset and an ageing standby circuit produce exactly the same outward behaviour. Do not conclude that a cable is loose until the latch pattern and the cold-start pattern have been checked.

"A bigger power supply will fix it." Wrong. If the cause is a short on the graphics card dragging the protection down, a bigger supply only feeds more current into the short. It does not fix the problem, and it carries a risk of burning something. Run the remove-the-card comparison first.

"A flat CMOS battery is why the machine will not start." Overstated. A flat battery normally only causes BIOS settings and the clock to be lost; the machine generally still powers on. Do not present the battery as the main cause of a power-on failure.

"Needing several presses means the card is dead." Wrong. Until the remove-the-card comparison has been done, no part may be declared the culprit. The comparison is what separates a card-side fault from a supply, motherboard or button fault.

"Shorting the power pins with a screwdriver is a standard test." This guide does not teach it. If the motherboard has its own power button, use that as the comparison. If it does not, the button and its cable are checked by reseating and by inspection, not by shorting pins.

When to Seek Inspection

Seek professional inspection if the machine still needs several presses with the card removed, if the fans spin once and stop even with the card removed, if the machine fails to start after the cables have been reseated and the button checked, or if there is any sign of physical damage inside the case. Bring the machine as it is, and note down what each press did, which steps were tried, and what each comparison showed. That record shortens the inspection considerably.

Inspection fees follow the published three-tier structure of HK$400–2,800, with the actual quote given after inspection. The workshop is at Room 1209, Dali Plaza, Shek Mun, Sha Tin. Telephone 2186 8010, WhatsApp 6558 6806. There is no on-site service.

FAQ

Pressing the button several times before the PC starts — does that mean the power supply is not powerful enough?

No. Wattage and starting are two different questions. A supply with a faulty standby circuit can fail to start the machine regardless of its labelled wattage, and a machine that fails to start does not prove the wattage is insufficient. The causes to check are the standby circuit, a latched protection circuit, a loose cable, the power button, and — through the remove-the-card comparison — a short on the graphics card. No generic wattage threshold is used as a standard in this guide.

The fans spin for a second and then stop — what does that mean?

It usually means the power supply protection has tripped, or a short on the load side has dragged the supply down. The supply switches on, detects the fault, and shuts itself down immediately to protect the machine. The next step is the remove-the-card comparison: if the machine starts reliably with the card removed, suspicion points to the card side; if it still spins once and stops, the fault is on the supply or motherboard side.

How do I do the remove-the-card test? If the PC boots with the card removed, does that mean the card is dead?

First switch the machine off, unplug the mains power cable, and wait for the residual charge to dissipate. Then remove the graphics card and try to start the machine — using integrated graphics if the machine has it, or simply watching whether it now starts reliably. If it starts reliably with the card removed, suspicion points to the card side, but that is not proof the card is dead: the cause of the short still has to be established by inspection. The comparison only shows which side the fault is on.

After a failure, pressing the button again immediately does nothing, but unplugging the power and waiting a while makes it start — what causes that?

A protection latch that has not reset. When the power supply trips its over-current or short-circuit protection, the latch stays on until the supply is fully disconnected from the mains. Pressing the button again while the supply is still connected to the wall does nothing, because the latch is still holding. Unplugging the mains cable and waiting for the residual charge to dissipate lets the latch reset, and the next press can then start the machine.

The PC is hard to start after sitting unused for a long time, but works normally once it is on — what is the problem?

This pattern points to an ageing standby circuit in the power supply. After a long time switched off, the standby circuit is slow to respond, so the first press misses its timing window and the Power Good signal arrives too late. Once the machine is running, the standby circuit is no longer involved, so it behaves normally. The pattern is diagnostic: it separates a standby circuit problem from a loose cable or a latched protection circuit.

Can a flat CMOS battery stop the PC from powering on?

Generally no. A flat CMOS battery normally only causes BIOS settings and the clock to be lost; the machine generally still powers on. It should not be treated as the main cause of a power-on failure. If the clock resets every time the mains is unplugged, the battery is worth replacing as a side check, but it does not explain a machine that needs several presses to start.

The several-presses problem only appeared after changing the CPU or RAM — where should I look?

Look first at whatever was touched during that change. A part change can reset BIOS settings or disturb a connector, and it is a trigger event rather than proof that the new part is faulty. Restore the state from before the change where possible, check the connectors that were moved, and run the remove-the-card comparison before declaring any part faulty. The guides on a GPU that is not detected after a CPU change and after a RAM change follow the same trigger-event logic.

I followed the whole decision tree and the PC still only starts intermittently — when should I seek inspection?

Seek inspection if the machine still needs several presses with the card removed, if the fans spin once and stop even with the card removed, or if the cables have been reseated and the button checked and the fault remains. Bring the machine as it is, with a note of what each press did and what each comparison showed. Inspection fees follow the published three-tier structure of HK$400–2,800, with the actual quote given after inspection.

Related articles

Related services

If you need Power connector repair, see these service pages: