GPU Shows No Signal When the PC Is Powered On Before the Monitor, but Works After a Restart or Replug? Power-On Order, Handshake Window & Restart-vs-Replug Diagnosis Guide

Key Takeaways
If a machine shows no signal when the PC is switched on before the monitor, yet shows a picture every time when the monitor is switched on first and allowed to wake fully before the PC is switched on, the order-flip comparison is the test that opens this guide: with the same machine, the same cable, and the same monitor, changing only the power-on order and watching the result reverse separates a handshake-window timing problem from a faulty card, a faulty cable, or a wrong input source, and this guide explains the handshake window, the restart-versus-replug comparison, and the further single-variable checks that confirm it.
The Order-Flip Test Comes First
This guide opens with a test, not with a conclusion. Take the machine exactly as it is, with the same cable and the same monitor, and run the power-on sequence in both orders. In the first round, switch the monitor on first, wait until it has fully woken and is showing its idle state, and only then switch the PC on. In the second round, switch the PC on first and switch the monitor on afterwards. Change nothing else between the rounds: no cable swap, no input change, no settings change, no second monitor added or removed. If the first order gives a picture every time and the second order gives no signal every time, the result has flipped with order alone, and that stable reversal is the defining shape of this guide. Everything else in this article exists to confirm that shape, to rule out lookalikes, and to explain what the reversal means and what it does not mean.
The reason the order test comes first is that it is the only test that isolates timing. A single no-signal event proves very little on its own. A card can fail to show a picture for many reasons, and most of them have nothing to do with power-on order. But when the same hardware produces a picture in one order and no picture in the other, repeatedly, the hardware itself has demonstrated that it can produce a picture. The difference between the two rounds is not the card, not the cable, and not the monitor, because none of those changed. The difference is the sequence in which two working parts were woken. That is why a stable reversal matters more than any single observation, and why this guide insists on repetition: run each order more than once, and only treat the pattern as established when the results stay opposite across the repeats.
What Exactly Counts as This Symptom
The symptom this guide addresses is narrow, and the narrowness is deliberate. The machine is switched on from a cold start. The PC is switched on before the monitor. The monitor, once it wakes, shows no signal, and the picture never arrives on its own, however long the machine is left alone. Switching the monitor on first, and letting it wake fully before the PC is switched on, gives a normal picture every time. When the fault state has occurred, either restarting the PC while the monitor stays on, or unplugging and replugging the display cable while both devices stay on, brings the picture back. That combination — order-dependent, self-sustaining until interrupted, and recoverable by restart or replug — is the whole symptom.
Several nearby shapes look similar at first glance and must not be folded into this one. If the picture eventually arrives on its own after a wait, that is a delay, not a deadlock, and it belongs with delayed-display diagnosis rather than this guide. If the monitor has jumped to a different input source, the picture is not missing; it is waiting on an input that is not selected, and that belongs with input-source diagnosis. If the picture is absent during startup but appears once the operating system has loaded, the handshake eventually succeeded at a later stage, which is a different timing story from a handshake that never completes. If the picture appears, but on the wrong monitor in a multi-monitor setup, that is a display-assignment question, not a no-signal question. If no order of powering on ever produces a picture, that is a general no-signal fault, and the order comparison cannot be the deciding test because there is no working order to compare against. Each of these neighbours has its own checks, and applying this guide's steps to them will mislead, because the steps assume a working order exists and a failing order exists side by side.
The symptom also has to be described in terms of what was actually done, not in terms of a presumed cause. Saying that the card is broken is a conclusion, and this guide does not allow that conclusion to be drawn from a single no-signal event. Saying that the cable is loose is also a conclusion, and the replug test below exists precisely because replugging changes more than one thing at once. The honest description is behavioural: in this order the picture appears, in that order it does not, and a restart or a replug restores it. Keeping the description behavioural is what makes the later steps meaningful, because each step then tests one alternative explanation against that behaviour.
The Mechanism: A Handshake Window, Not a Speed Contest
When a PC starts, the graphics card and the monitor must complete a mutual confirmation before any picture can appear. The card needs to learn what display is connected and which display modes that display supports, and the monitor needs to be in a state where it can receive that enquiry and answer it. This confirmation is sometimes called a handshake, and the word is useful because a handshake needs both parties ready at the same time. A handshake offered to an empty room is not a failed handshake in the sense of a broken part; it is an enquiry that nobody was present to answer.
When the PC is switched on before the monitor, the card may send its confirmation enquiry while the monitor is still waking. If the monitor is not yet ready to receive and answer, the enquiry goes unanswered. The card, receiving no answer, may proceed on the assumption that no display is connected, and it may not spontaneously ask again. The result is a one-sided deadlock. The computer behaves as though no monitor exists, and the monitor, once it has finished waking, receives no fresh enquiry, so it keeps showing no signal. Neither side is broken. Each side is waiting for the other to make a move that neither side has been told to make again. That is why waiting does not resolve this shape: there is no pending process that will finish on its own. The deadlock persists until something forces a fresh enquiry.
The decisive concept here is a window rather than speed. The problem is not that one side is slow in some absolute sense, and it is not that one side is too fast. The problem is that the periods during which each side is ready to confirm do not overlap. Switching the monitor on first moves its readiness earlier, so that by the time the card sends its enquiry, the monitor is already ready to answer, and the window overlaps. The same two devices, with the same capabilities, succeed or fail depending purely on whether their readiness windows coincide. That is also why the symptom can feel so inconsistent to the person experiencing it: on mornings when the monitor happens to be switched on first, everything works, and on mornings when the PC happens to be switched on first, nothing works, and the difference can look like randomness until the order is noticed and controlled.
Two further points about the mechanism keep the diagnosis honest. The first is that a restart and a replug, which both restore the picture, do so by the same route: they force a fresh confirmation. A restart asks the card to begin its startup sequence again, this time while the monitor is already awake and ready, so the enquiry is answered. A replug tells both sides that a connection has newly appeared, which triggers a fresh enquiry on a connection that is now ready at both ends. Neither action repairs hardware, and neither action proves that hardware was faulty. They are both ways of breaking a deadlock by starting the conversation again. The second point is that the mechanism says nothing about which interface is involved. The handshake idea applies wherever a card and a monitor must confirm a connection at startup, and this guide deliberately avoids ranking interfaces against each other. What matters is the order relationship measured on the machine in front of you, not a general claim about one interface being better than another.
Step One: The Order Comparison, Done Strictly
The order comparison has already been introduced, and here it is specified strictly enough to be trusted. Use the same machine, the same cable, the same monitor, and the same monitor settings throughout. In the monitor-first round, switch the monitor on, wait until it has fully woken and settled, and then switch the PC on. In the PC-first round, switch the PC on and switch the monitor on afterwards, in the way that previously produced the fault. Repeat each round more than once. Record the result of each round in writing, as a simple pair: monitor-first gave a picture, PC-first gave no signal, or whatever the actual results were.
Both outcomes carry meaning, and both must be read. If the results reverse stably — picture in the monitor-first order, no signal in the PC-first order, repeatedly — the handshake-window explanation is supported, and the later steps exist to rule out the remaining alternatives rather than to re-prove the reversal. If both orders give a picture, the machine does not currently show this symptom, and the fault, if it returns, needs to be observed again under controlled conditions before any conclusion is drawn. If both orders give no signal, the fault is not order-dependent at all, and general no-signal diagnosis is the correct path, not this guide. If the results vary from round to round without a stable pattern, the comparison is not yet controlled: something else is changing between rounds, such as a monitor setting, a different wake state, or a second display, and that uncontrolled variable must be found and held constant before the comparison is judged. A mixed or unstable result is not a failed test; it is information that the test conditions are not yet clean.
Step Two: The Restart Comparison
Once the fault state has been produced by the PC-first order, leave the monitor switched on and restart the PC alone, without touching the cable and without switching the monitor off and on again. This is a single-variable change: the only thing that changes is that the card runs its startup sequence again, this time with the monitor already awake.
If the picture appears after the restart, the result supports the handshake-window explanation. The card, asking again while the monitor is ready, receives the answer it missed the first time. The restart did not replace a part, did not reseat a connector, and did not change a setting, so the recovery points at timing rather than at hardware condition. If the picture does not appear after the restart, the explanation is less clear. A failed restart does not by itself disprove the handshake idea, but it does mean that a fresh enquiry under ready conditions was not enough on that occasion, and the investigation should move to the replug comparison and the later checks before any conclusion is drawn. Either way, the restart result should be recorded alongside the order result, because the pair together is stronger evidence than either alone.
Step Three: The Replug Comparison
The replug comparison is an alternative to the restart comparison, not an addition to be run at the same time. Starting again from the fault state, with both devices left switched on, unplug the display cable and plug it back in, and nothing else. Do not restart the PC as part of this step, and do not switch the monitor off and on again, because doing so would change more than one variable and blur the result.
If the picture appears after the replug, the result supports the same explanation by a different route: the fresh connection triggered a fresh confirmation, and this time both sides were ready. If the picture does not appear after the replug, the fault state is more stubborn than a single missed enquiry, and the later steps matter more. The important discipline in this step is about interpretation. A recovery after replugging does not, on its own, prove that the connector was loose. Replugging does two things at once: it remakes the physical contact, and it triggers a fresh confirmation. Only the order comparison, which changes no physical contact at all, can separate those two explanations. That is why the replug result must always be read together with the order result, never alone. A machine whose order comparison reverses stably and whose replug also recovers is showing a timing story; a machine whose order comparison does not reverse but whose replug recovers is showing a contact story, and the two stories lead to different next steps.
Restart and replug should not be ranked against each other. This guide does not claim that one is more effective than the other, and the reader should not treat a recovery by one route as stronger evidence than a recovery by the other. Both are ways of forcing a fresh confirmation, and either one, used alone as a controlled comparison, serves as evidence. Using both in sequence on the same fault state, without resetting to the fault state between them, would waste the comparison, because after the first recovery there is no fault left for the second to test.
Step Four: The Input-Source Exclusion
Before any deeper conclusion is drawn, check the monitor's selected input while the fault is present. A monitor that has moved to a different input source will show no signal on the selected input even though the card and the connection are behaving normally. The check is simple: confirm that the monitor is set to the input that the cable is actually connected to.
If the monitor is on the wrong input, the fault belongs with input-source diagnosis, not with this guide, and the order comparison should be repeated only after the input has been corrected and held constant. If the monitor is on the correct input and still shows no signal in the PC-first order, the input explanation is excluded, and the handshake-window explanation survives one more challenge. This step is placed after the restart and replug comparisons deliberately, because an input error can coexist with a timing problem, and the order of checking does not change the need to check it. What matters is that it is checked and recorded, not assumed.
Step Five: The Cable and Port Comparison
With the order relationship established, test whether the cable or the port is a hidden variable. Replace the cable with another cable of the same type, keeping the same order comparison, and repeat both orders. Then, if the machine offers another port of the same type, repeat the order comparison through that port as well. Change only one of these at a time, and keep the order rounds identical in every other respect.
If the order relationship survives the cable change and the port change — monitor-first still gives a picture, PC-first still gives no signal — then the cable and the port are not the main explanation, because the pattern persisted while they changed. If the order relationship disappears after a cable or port change — both orders now give a picture — then the cable or the port was a contributing factor, and the original cable or port deserves closer attention on its own terms, as a contact or compatibility question rather than a timing question. If the pattern changes in some other way, such as becoming unstable, the change itself is information, and the comparison should be repeated to see whether the new pattern is stable before it is interpreted. The point of this step is not to find a cable that works; it is to learn whether the order effect depends on the cable or the port at all.
Step Six: The Single-Screen Simplification
Where more than one monitor is connected, repeat the order comparison with only the affected monitor connected, and the other monitors disconnected. Multiple displays can interfere with each other's confirmation sequence, and that interference can mask or mimic an order effect. Simplifying to a single screen removes that interference as a variable.
If the order reversal is still visible with a single screen, the pattern is established cleanly, and the other monitors can be reconnected afterwards with the knowledge that the core effect does not depend on them. If the order reversal disappears with a single screen, the multi-monitor arrangement was part of the story, and the diagnosis shifts toward how the displays interact at startup, rather than toward the single-monitor timing story. If only one monitor was ever connected, this step is already satisfied, and it should be noted as such rather than skipped silently, because the record of what was tested matters as much as the record of what was found.
| Comparison | What the result proves | What it does not prove |
|---|---|---|
| Monitor-first gives a picture, PC-first gives no signal, repeatedly | The result follows power-on order, with hardware held constant | Which side missed the window, without further testing |
| Both orders give a picture | This symptom is not currently present | That the fault will never return, without repeated observation |
| Both orders give no signal | The fault is not order-dependent | What the underlying fault is, without general no-signal diagnosis |
| Results vary between rounds without a stable pattern | Test conditions are not yet controlled | Any conclusion about timing, until the uncontrolled variable is found |
| Restart with the monitor on restores the picture | A fresh enquiry under ready conditions succeeds | That hardware was faulty, since nothing was replaced or reseated |
| Restart with the monitor on does not restore the picture | A fresh enquiry alone was not enough on that occasion | That the handshake explanation is disproved, without the replug and later checks |
| Replug restores the picture | A fresh connection triggered a fresh confirmation | That the connector was loose, since replugging also retriggers confirmation |
| Replug does not restore the picture | The fault state is more stubborn than a single missed enquiry | That timing plays no part, without the full set of comparisons |
| Monitor is on the wrong input during the fault | The picture is waiting on an unselected input | Anything about timing, until the input is corrected and the order test repeated |
| Monitor is on the correct input and still shows no signal | The input explanation is excluded | That timing is the only remaining explanation, without the cable, port, and single-screen checks |
| Order relationship survives a cable and port change | The cable and the port are not the main explanation | That the cable and port are perfect in every respect, only that they do not explain the order effect |
| Order relationship disappears after a cable or port change | The cable or port was a contributing factor | That timing plays no part in the original setup, without retesting the original parts |
| Order reversal persists on a single screen | The effect does not depend on multi-monitor interference | How the displays interact when reconnected, which needs its own test |
| Order reversal disappears on a single screen | The multi-monitor arrangement was part of the story | That any single display is faulty, without testing each display alone |
Settings That Change the Comparison
Two settings deserve explicit attention, because changing them midway through testing invalidates the comparison. The first is the monitor's automatic input detection, where the monitor scans its inputs on its own rather than staying on a selected input. Automatic detection changes the timing with which the monitor becomes ready to answer, so a machine tested with detection on and a machine tested with detection off are not the same test. The rule is simple: keep the setting the same across every round of the comparison, and record what it was. Whether detection is on or off is a matter for the reader's own preference afterwards; during diagnosis, consistency is what matters.
The second is the difference between a cold start and waking from sleep. A machine that has been fully shut down and a machine that has been woken from sleep are not in the same state, and the startup confirmation sequence may differ between them. The order comparison should therefore compare cold start against cold start, and any wake-from-sleep observation should be recorded separately, not mixed into the same comparison. Keeping the startup method constant is a small discipline that prevents a large confusion, because a pattern that appears in cold starts and disappears on wake, or the reverse, is a statement about startup method, not about power-on order.
A related caution concerns interfaces. Different interfaces can differ in how strictly their confirmation sequence is handled, but this guide does not turn that observation into a general ranking. A reader who finds an order effect on one interface should report it as a finding about their own machine, not as a claim about that interface everywhere. The diagnostic value lies in the measured order relationship on the machine being tested, and a ranking table would add claims the tests in this guide never made.
Fast startup deserves a sentence of its own. A fast-startup setting changes what a shutdown and a boot actually mean, so a machine that was only apparently fully shut down cannot automatically be counted as a cold start. Hold the startup method constant across every comparison, and do not mix a fast-startup boot with a genuinely cold boot.
Boundary Cases: When It Is Not This Symptom
A picture that eventually arrives on its own after a wait is a delay, and it should be judged as a delay, with its own questions about what finishes late and why. A monitor that has jumped to a different input is an input-selection matter. A picture that is absent during startup but appears once the operating system has loaded is a later-stage success, not a deadlock. A picture that appears on the wrong monitor is an assignment matter. A machine that never shows a picture in any order is a general no-signal fault. A machine whose fault appears only with several monitors connected is a multi-display interaction matter until proven otherwise. A machine whose fault changes when the monitor's automatic detection setting is changed is a settings-interaction matter, and the comparison must be redone with the setting held constant before any timing conclusion is drawn.
There is also a boundary around what this guide does not claim. It does not claim that any particular card, monitor, cable, or interface behaves this way in general. It does not claim that switching the monitor on first is a repair. It is a habit adjustment that avoids the deadlock by aligning the readiness windows, and it is a perfectly reasonable everyday workaround, but it does not change the underlying timing behaviour of the devices. If the reader wants the underlying behaviour investigated further, the recorded comparisons in this guide are the evidence to bring, not a claim that a part has failed.
Common Misconceptions
Misconception one: a single no-signal event means the graphics card has failed. It does not. A card that produces a picture every time in the monitor-first order has demonstrated that it can produce a picture. Judging the card from one order alone ignores the working order entirely.
Misconception two: a faulty cable would behave this way. A faulty cable usually fails regardless of power-on order, because a broken connection does not repair itself when the sequence changes. When order alone stably reverses the result, the cable is not the main explanation, and the cable-and-port comparison exists to test that point rather than assume it.
Misconception three: waiting longer will eventually bring the picture back. In a delay, waiting helps. In a deadlock, waiting does not, because nothing is in progress that will finish on its own. The distinction between waiting and intervening is the distinction between the delay shape and this shape, and confusing the two leads to long, fruitless waits.
Misconception four: recovering after a replug proves the connector was loose. Replugging does two things at once, remaking contact and retriggering confirmation. The recovery is consistent with a contact problem, and it is equally consistent with a timing problem, and only the order comparison, which changes no contact at all, separates the two.
Misconception five: switching the monitor on first is unnecessary fuss, or switching the PC on first is a bad habit that deserves blame. Neither judgement helps. Power-on habits are formed for sensible reasons, including saving power, and the point of this guide is not to criticise a habit but to show that, on a machine with this timing behaviour, order is a variable with real consequences. Adjusting the habit is a practical response, not a moral one, and it costs nothing to try.
What to Record, What to Bring, and When to Stop
Record the order comparison as a written pair, round by round, with the monitor setting and the startup method noted alongside. Record the restart result and the replug result separately, each from its own fresh fault state, not from a chain in which one recovery followed another. Record the input check, the cable and port comparison, and, where relevant, the single-screen comparison, each with its before-and-after result. That record is the whole diagnosis in miniature: anyone reading it can see which variables were changed, which were held constant, and which explanations survived.
Bring that record to any inspection, together with a note of the monitor's automatic detection setting, the startup method used, and whether the pattern is stable or intermittent. Describing the two orders and their opposite results is enough for a quick judgement on the next step, and it avoids repeating tests that have already been done carefully.
Readers nearby are welcome to bring their records to HK GPU Fix for advice, at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806. There is no Mong Kok shop, and there is no home-visit service.
Stop changing parts or settings if any of the following appears. The machine never shows a picture in any order, which points to general no-signal diagnosis rather than this guide. The picture shows corruption, flickering, or artefacts once it appears, which are failures beyond a handshake deadlock. The pattern changes unpredictably even with every variable held constant, which suggests an unstable fault that needs hands-on inspection rather than further home testing. Or any step feels unsafe, in which case the step should be left undone and advice sought first. Stopping at that point is not giving up; it is recognising that the evidence has reached the edge of what order comparisons can decide.
Frequently Asked Questions
Why does switching the monitor on first fix the problem, if nothing is broken?
Because it aligns the readiness windows. The monitor is already awake and ready to answer when the card sends its confirmation enquiry, so the handshake completes. Nothing was repaired, because nothing was broken; the sequence was changed so that both sides were ready at the same time.
Why does the picture never arrive on its own in the failing order?
Because the deadlock is stable, not pending. The card asked while the monitor was not ready, received no answer, and did not ask again. The monitor, once awake, receives no fresh enquiry. Without an intervention that forces a fresh enquiry, neither side has a reason to move.
Does restarting the PC prove the card is faulty?
No. A restart forces a fresh startup sequence while the monitor is already awake, so it tests timing, not hardware condition. No part was replaced or reseated during a restart, so a recovery after restart points at the sequence, not at a failed part.
Does recovering after replugging prove the cable is loose?
No, not on its own. Replugging remakes the physical contact and also triggers a fresh confirmation. Only the order comparison, which changes no contact, can separate a contact explanation from a timing explanation. Read the replug result together with the order result, never alone.
Should I use restart or replug as my test?
Either one, used alone from a fresh fault state, serves as a comparison. Both force a fresh confirmation by different routes. Do not rank them, and do not run one immediately after the other on the same fault state, because after the first recovery there is no fault left for the second to test.
Why does the same machine work on some days and fail on others?
Most likely because the power-on order varied between those days without being noticed. On days when the monitor happened to be switched on first, the windows overlapped and the picture appeared. On days when the PC happened to be switched on first, they did not. Controlling the order turns apparent randomness into a stable pattern.
Is switching the monitor on first a repair or just a workaround?
It is a practical habit adjustment that avoids the deadlock. It does not change the underlying timing behaviour of the devices, and it should be described as such. If the underlying behaviour needs further investigation, the recorded order comparisons are the evidence to bring.
What should I check first if the fault appears?
Check the monitor's selected input first, so that an input error is excluded, then run the strict order comparison with everything else held constant. A single no-signal event, without the comparison, does not support any conclusion about the card, the cable, or the monitor.
What should I bring if I seek advice?
Bring the order comparison round by round, the restart result and the replug result each from its own fresh fault state, the input check, the cable and port comparison, and, where relevant, the single-screen comparison, together with the monitor's automatic detection setting and the startup method used. That record lets advice start from the comparisons rather than repeating them.