Desktop Colours Wrong After Exiting a Game: Game Colour Settings, ICC Profile and Fullscreen Mode Diagnosis Guide

2026-11-19 | GPU Repair | 51 minutes | HK GPU Fix Editorial Team

Desktop Colours Wrong After Exiting a Game: Game Colour Settings, ICC Profile and Fullscreen Mode Diagnosis Guide

Key Takeaway

Your desktop colours were normal before you played, but after you quit the game they look washed out, stronger, or tinted. Check the same picture before and after playing: if the colour changed after quitting, the game triggered it; then check whether your profile is still loaded. That pair of checks separates a game fault from a monitor fault.

The Shape of the Problem: Normal Before Play, Wrong After Quitting

This guide covers one single narrow pattern, and the diagnosis never leaves that pattern. Before the game is opened, the desktop colours look normal. The game itself, while it is running, looks normal too. Then the game is closed, the desktop comes back, and its colours have changed: washed out, stronger, tinted, or brighter than they were before the game was opened. A restart, or applying the colour profile again, brings the colours back to normal. The machine still reaches the desktop, it never shows a blue screen, and it does not restart on its own. Only the colours have moved, and they moved when the game was closed and at no other time.

Three conditions pin the shape down, and each one has to hold. First, the trigger is quitting a game: the desktop colours were normal before the game was opened, and they are wrong after it is closed. Second, the comparison is made on one machine, on one screen, with the same picture — the same desktop wallpaper or the same photograph — looked at before play and after quitting. Third, the colours inside the game itself were normal while it was running. The fault shows up only after quitting, on the desktop.

That boundary matters, because readers usually treat “the colours changed” as a finished answer. It is not. A colour change after quitting a game is not one thing. It can be a game that changed the graphics card’s correction table and did not put it back, a colour profile that was swapped or lost, a display mode that handles colour unlike another mode, a high-dynamic-range setting that was left on, a night mode that happened to switch on at the same moment, or a monitor maker’s tool that reset the profile in the background. Each one calls for a test of its own. The question to answer is not why the colours changed in general. It is which of those actors changed the colours, and which one failed to put them back — and the answer may sit in the game’s exit routine, in the profile association, in the display mode, in a leftover high-dynamic-range state, or in a tool running in the background. Only the before-and-after comparison pulls the first answer clear of the rest.

One warning belongs before any check runs. “The colours changed” is what the eye reports, not a measurement. No two people describe the same change in the same words, and a change that looks large to one person can look small to another. That is why every check in this guide compares the same picture, looked at before play and after quitting, on the same screen. The comparison is the evidence. A description on its own — “washed out”, “stronger”, “tinted” — is not a measurement, and it cannot take the place of one. The baseline picture is what every later check is weighed against, and without it, a later look cannot be judged a change or no change — there is nothing to weigh the later look against.

One more boundary belongs before any check runs. A wide-gamut screen showing standard colours too strongly is not taught here. One move is used from that topic — noting whether the colours were already too strong while the game was running — and nothing further. Two screens showing the same picture in unlike colours is not taught here either. The one move taken is to note whether the change appears on one screen or on both — and nothing further. High dynamic range on its own is not taught here either. The one move taken is to look at whether the high-dynamic-range switch is still on after quitting — and nothing further. Colour calibration by itself is not taught here either. The one move taken is to open Colour Management and read which profile is set as the default — and nothing further. Using one move from each nearby topic is where the piece stops; those topics themselves are left alone.

Why Quitting a Game Can Change the Desktop Colours

Two things must be kept apart before anything else, because confusing them is the commonest error in this diagnosis. The first is the ICC profile: a file that states how a screen turns digital red, green, and blue numbers into actual light output. The second is the calibration loading: a correction table, called a gamma ramp, that is loaded into the graphics card and genuinely alters the output. Windows can have a profile selected as the default without its calibration having been loaded, because the two are managed as separate stages. Published guides put it plainly: choosing a profile does not equal loading its calibration. So a profile can still be sitting exactly where it was, untouched, while the colours have changed — because the correction table was altered and not put back. A reader who opens Colour Management, sees the profile still there, and concludes "it has nothing to do with the profile" has mistaken one thing for the other. The profile and the calibration must be checked as two separate things, and the checks below do exactly that.

The game itself supplies the second mechanism. In exclusive full-screen mode, a game can change the graphics card's gamma table directly — adjusting brightness or contrast for its own picture — and the normal routine is to put the table back to the desktop setting when the game is closed. Some games do not put it back on exit. The desktop is then left sitting in the game's gamma, and its colours look changed: washed out, stronger, tinted, or brighter than before. Published guides describe this as a known shape: after entering a full-screen game, the picture calibration resets, the brightness changes, and the profile stops taking effect. The fault sits in the game's exit routine, and it is not a verdict on the screen or on the driver. A screen that shows wrong colours because a game left its gamma behind is a screen obeying the table it was given, not a screen that has failed.

Game Bar and night mode supply a third mechanism, and it is a matter of timing. Game Bar, in older versions from the Windows 10 Creators Update era, interfered with picture calibration and with ICC profiles. Night mode changes the colour temperature on a schedule, making colours look warmer. Either one can happen to act at the very moment a game is closed, and the reader then blames the game for a change the game did not make. The two must be pulled apart by a time comparison: night mode has a scheduled time, and Game Bar can be switched off one program at a time. A coincidence in time is not a cause: the two arriving together proves nothing about which one acted. The Game Bar position also needs a plain statement about age: the published account of Game Bar interfering with calibration comes from the Creators Update era, and Game Bar today is a different version. An old account cannot be carried forward as a conclusion about the current version. It can only be tested, by switching Game Bar off and playing again.

High dynamic range supplies a fourth mechanism. A game that runs in high dynamic range can leave Windows still in its high-dynamic-range container after quitting, with the desktop continuing to run inside it, and standard content then looks washed out or grey. Published guides state that Windows 11, with high dynamic range on, forces desktop programs into the high-dynamic-range container, and that on a screen whose peak brightness is not high, desktop colours look washed out and grey. This layer sits beside the guide on high dynamic range, which treats that topic in full; this piece uses only the high-dynamic-range switch comparison after quitting, and does not redo that guide. Two shapes must be kept apart here as well: colours wrong while high dynamic range is on, and colours wrong after quitting because high dynamic range was left on. They are unlike shapes, and only the second one belongs to this guide.

A monitor maker’s tool supplies the fifth mechanism. Some monitor tools have an automatic mode that keeps resetting the external screen’s profile back to the one the tool itself chooses. Published guides describe cases where a user had to switch the tool to manual mode before the resetting stopped. That mechanism explains a shape readers often describe as baffling: the profile was set correctly a moment ago, the game was opened once, and the profile has changed back again. The game did not change it. The tool in the background changed it. This layer stands on its own as a separate suspicion, never lumped together with the game’s gamma, because the two have different tests and different follow-up.

Five Comparisons That Narrow Things Down

The diagnosis turns on five comparisons, and each one ends in one of two directions. They belong in the order printed: each answer leaves the next comparison less to sort out, and the opening comparison — the same picture before play and after quitting — counts for more than any check that follows. No single reading on its own decides anything; what decides is which reading moves when a single variable changes, in one run, on one machine.

The first test is the before-and-after comparison, and it matters most in this piece. The same picture — the same desktop wallpaper or the same photograph — is looked at on the same screen before the game is opened, and looked at again immediately after the game is closed. If the colours changed only after quitting, the game triggered it, and the diagnosis stays here. If the colours were already changed before the game was opened, the game is not the trigger, and the reader should turn to the guides for wide-gamut colour, night mode, or backlight ageing. The strength of this test is that it moves one variable only — whether the game has been opened and closed — and leaves the rest untouched. The comparison must be made against the same picture, because a different picture cannot be weighed against the first one.

The second test puts the profile against the calibration. After quitting, Colour Management is opened and the default profile is read. Two results are possible, and they lead in opposite directions. The profile is still the original one, but the colours have changed: the calibration-loading layer is the suspect, because the gamma table was altered and not put back while the profile itself never moved. The profile has changed to another one, or the original one is gone: the association layer is the suspect, because the profile itself was swapped or lost. This fork decides which way the diagnosis goes from here; leaving it out is not an option. Published guides state the split plainly: a profile still selected with colours changed, and a profile gone, are two unlike problems, and each must be checked on its own.

The third test is the display-mode comparison. The same game gets one run in exclusive full screen and a second run as a borderless window. Wrong only after exclusive full screen, and normal after a borderless run, points at the full-screen mode layer: the exclusive mode can change the graphics card's gamma table directly, and the game must put it back on exit. Wrong after both modes means the mode is not the layer, and the hunt returns to the game, the driver, and the tools. This test uses the display-mode guide's mechanism, and does not rewrite that guide.

The fourth test is the high-dynamic-range switch comparison. After quitting, the Windows display settings are opened and the high-dynamic-range switch is read. Still on after quitting points at the high-dynamic-range leftover layer: the desktop is still sitting inside the high-dynamic-range container, and standard content looks washed out or grey. Never switched on in the game at all points away from high dynamic range, and the diagnosis goes back to the gamma and profile layers. The switch is the evidence, not a description of how the colours look, because a washed-out look on its own cannot tell a high-dynamic-range leftover apart from a gamma table left behind.

The fifth test sets one suspect program against the rest. The game itself, Game Bar, the monitor maker's tool, night mode, and the calibration software are switched off one at a time, and the game is opened and closed again after each one. Whichever program, once switched off, stops the trouble from coming back is the one at fault. Turning them all off together proves nothing, because the result then carries no name — the test cannot point at any one program. Published guides state that a monitor tool's automatic mode can keep resetting an external screen's profile, and that Game Bar in older versions interfered with calibration — and both of those are claims to test one program at a time, not conclusions to adopt in advance.

TestWhat is comparedOutcome pointing at the game or profileResult pointing elsewhere
Before play against after quittingSame picture on the same screen, before the game and after quittingColours changed only after quitting; the game triggered itAlready changed before the game; not the game
Profile still there against profile goneDefault profile read in Colour Management after quittingProfile still the original one but colours changed; calibration-loading layerProfile changed to another one, or gone; association layer
Exclusive run against borderless runSame game, each display mode in turnWrong only after the exclusive run; full-screen mode layerWrong after both modes; not the display mode
High-dynamic-range switch after quittingWindows display settings read after quittingSwitch still on; high-dynamic-range leftover layerNever switched on in the game; not high dynamic range
One suspect program at a timeGame Bar, monitor tool, calibration software, night mode switched off singlyOne program switched off stops the trouble; that program is the actorNo single program stops it; the game's own exit routine

How Profiles, Calibration, and Display Modes Interact

The profile and the calibration are worth a closer look, because they are the pair readers most often collapse into one thing. An ICC profile is a file. It states how a screen turns digital numbers into actual light. It does not, by itself, change what the screen shows. The calibration — the gamma ramp — is a correction table loaded into the graphics card, and it is the table that genuinely alters the output. Windows manages the two as separate stages: a profile can be selected as the default, and its calibration can still fail to load, or can be overwritten afterwards by a game, a tool, or a driver setting. That separation is the whole reason the second test reads the profile and the colours as two separate results. A profile still sitting in place with colours changed is not a contradiction. It is the expected shape of a calibration that was altered and not put back.

The game’s gamma mechanism needs the same care. Under exclusive full-screen mode, the game can write to the graphics card’s gamma table, and it can adjust brightness or contrast for its own picture. The normal routine is to restore the table when the game closes. Some games do not restore it. The desktop is then left in the game’s gamma, and the colours look changed until a restart, or a fresh application of the profile, reloads the table. That is why restarting the machine, or applying the profile again, brings the colours back: the reload restores the table, and the restoration proves nothing about the screen. It proves only that the table was the thing that had changed. A screen that returns to normal after a restart is a screen that was obeying a wrong table, not a screen that has healed itself.

The display mode decides how much of the output the game can reach. In exclusive full screen the game can write to the gamma table itself, as part of its own picture, while borderless windowed play keeps the desktop composition in charge and the game cannot reach the table the same way. That is why the same game can leave the colours wrong after an exclusive run and leave them untouched after a borderless run, and why the mode comparison is a real test rather than a preference. The mode is not a repair: playing borderless avoids the shape, and it does not fix the game’s exit routine. The mode comparison names the layer, and the naming is the work.

High dynamic range and night mode both need the same discipline about timing. High dynamic range left on after quitting is a leftover state, and it is read from the switch, not from the look of the colours. Night mode arriving on schedule at the same moment a game is closed is a coincidence in time, and it is read from the schedule, not from the look of the colours. Both shapes look like a game that changed the colours, and both are pulled apart by reading the setting rather than the picture. A washed-out desktop with the high-dynamic-range switch still on is a leftover state. A warm desktop with night mode's scheduled time matching the moment of quitting is a schedule coincidence. The setting is the evidence in both cases, and the description of the colours is not.

One more connection is worth attention: how the profile, the tool, and the driver act on each other. A monitor maker’s tool in automatic mode can reset the profile in the background, a driver colour setting can overwrite the calibration, and a calibration program can reload its own table on a schedule. Any one of them can act while a game is running, and the reader then blames the game for a change made by a program that was running all along. That is why the fifth test takes the suspect programs off singly: when switching off no single program stops the trouble, the game’s own exit routine is what remains, not a tool. The follow-up there is not to blame all the tools at once, but to leave on only the programs the owner really uses, turn the rest off, and try the smaller set again.

The Five Suspicion Layers

Any case of desktop colours shifting after a game is closed sits inside one of the five layers below. They appear in the order the evidence most often hints at first, but that order proves nothing by itself: a layer is named only once a test has supported it or ruled it out. The published sources part ways on which layer appears most often, and this guide puts none of them ahead of the rest. Every layer below lists what would point towards it; the evidence decides.

LayerWhat it coversEvidence pointing at it
First layer: game gamma not restored layerA full-screen game changed the gamma table and did not put it back on exitWrong only after exclusive full screen; profile still the original one but colours changed; applying the profile again or restarting brings the colours back
Second layer: profile association layerThe profile itself was swapped for another one, or is goneDefault profile in Colour Management has changed to another one, or the original one is gone; it must be added again before the colours return
Third layer: full-screen mode layerIn exclusive full screen the game can write to the gamma table; in a borderless window it cannotWrong only after an exclusive run; normal after a borderless run
Fourth layer: high-dynamic-range and night-mode leftover layerHigh dynamic range left on after quitting, or night mode arriving on schedule at the same momentWindows high-dynamic-range switch still on after quitting; night-mode scheduled time matches the moment of quitting
Fifth layer: driver and monitor-tool overwrite layerA driver colour setting, a monitor maker's tool, or calibration software changed the profile in the backgroundTrouble stops after one tool is switched off; the profile changes back again after a while

The first layer is the game gamma not restored layer, and it is the layer the before-and-after comparison is built to find. A full-screen game changed the gamma table and did not put it back when it closed. The evidence is the first test itself: the colours were normal before play and wrong after quitting, and the profile is still the original one while the colours have changed. Supporting evidence is that applying the profile again, or restarting the machine, brings the colours back at once — both reload the table. The follow-up is to record the game and the display mode it ran in, since that record names the layer, and an inspection builds on records of that kind. This layer is not a broken screen. A screen showing wrong colours because a game left its gamma behind is a screen obeying the table it was given, not a screen that has failed.

The second layer is the profile association layer. The profile itself was swapped for another one, or is gone altogether. The evidence is read in Colour Management after quitting: the default profile has changed to another one, or the original one is gone and must be added again before the colours return. This layer sits beside the guides on wide-gamut colour and on two-screen calibration, which treat profile selection in full; this piece uses only the profile reading, and does not redo their calibration work. The follow-up is to add the original profile back and set it as the default, and then to run the one-at-a-time test, because a profile that keeps changing back again points at a tool changing it in the background, not at the profile itself.

The third layer is the full-screen mode layer. In exclusive full screen the game can write to the gamma table, while in a borderless window the desktop composition stays in charge, so the two modes handle colour unlike each other. The evidence is that the colours go wrong only after an exclusive run, and stay normal after a borderless run. This layer uses the display-mode mechanism from the display-mode guide, and does not redo the guide itself. The follow-up is to record which mode stayed normal and which went wrong, because that record names the layer. Playing borderless avoids the shape; it does not fix the game’s exit routine — borderless play is not a repair.

Fourth in the list is the high-dynamic-range and night-mode leftover layer. High dynamic range can be left on after quitting, leaving the desktop inside the high-dynamic-range container, and night mode can arrive on its schedule at the very moment a game is closed. The evidence is read from the settings, not from the look of the colours: the Windows high-dynamic-range switch still on after quitting points at the leftover layer, and a night-mode scheduled time matching the moment of quitting points at a coincidence in time. Nothing in this layer is a fix: the switch is read, the schedule is read, and the outcome is noted down. A switch left on shows the trouble sits in the leftover layer. It does not show the game was innocent, because the gamma layer and the leftover layer can exist at the same time.

The fifth layer is the driver and monitor-tool overwrite layer, and it is an exclusion layer. A driver colour setting, a monitor maker’s tool in automatic mode, or calibration software can change the profile in the background, and the reader then blames the game for a change made by a program that was running all along. The evidence is that the trouble stops after one tool is switched off, or that the profile changes back again after a while on its own. This layer never names a tool as the culprit outright. The test to run is to separate the tools one at a time. The follow-up is to switch the suspect programs off one at a time, and to switch a monitor tool from automatic mode to manual mode as a test where the published guides describe that step.

The Seven Diagnostic Steps

Take the seven checks in the order printed below. Each check's outcome decides which check follows it. Trying a fresh calibration, or a driver reinstall, before the first three checks have run usually throws the attempt away: no baseline in hand, no before-and-after comparison run, and no profile reading taken yet, a later change proves nothing, even if the colours come back.

CheckWhat to doHow the result reads
Check 1Pin down the baseline and record the colours: same wallpaper or photograph, same screen; note the default profile name, the high-dynamic-range switch, and the night-mode switch before playWith no baseline, no later look means anything; a description from memory is not a record
Check 2Run the before-and-after comparison, the deciding check: open the game, play a while, quit back to the desktop, and look at the same picture at onceColours changed means the game triggered it; no change means it is not the game
Check 3Read the profile: open Colour Management after quitting and read the default profileChanged to another one, or gone, means the association layer; still the original one but colours changed means the calibration-loading layer
Check 4Run the full-screen against borderless comparison: the same game, first exclusive full screen, then a borderless windowWrong only after exclusive full screen means the full-screen mode layer; wrong after both means it is not the mode
Check 5Run the high-dynamic-range and night-mode comparison: read the Windows high-dynamic-range switch after quitting, and read the night-mode scheduleSwitch still on means the high-dynamic-range leftover layer; schedule matching the moment of quitting means a coincidence in time, not the game
Check 6Switch the suspect programs off one at a time: Game Bar, the monitor maker's tool, calibration software, each off singly, then open and close the game againWhichever program, once off, stops the trouble names the layer; all off at once finds nothing
Check 7Halt line: the profile is gone, applying it again does not bring the colours back, and a broken picture, a black screen, or a crash has turned up beside the colour changeBring the records — the before-and-after comparison result, the profile name, the high-dynamic-range switch, and the checks run so far — to bring to the inspection; do not delete and install profiles one by one, and do not download a profile for an unknown screen from the internet

The first check underpins every check after it, and it runs before any test begins. Pin down the baseline: the same desktop wallpaper or the same photograph, on the same screen, looked at before the game is opened. Note down the colours as they look, and note the settings beside them: the default profile name in Colour Management, the high-dynamic-range switch, and the night-mode switch. A photograph of the screen is a record; a description from memory is not. Every check after this one is weighed against the baseline. Without a baseline, a later look cannot be judged a change or no change — there is nothing to weigh the later look against. Where the colours from before the trouble started were never noted down, say so plainly: a remembered impression is not a measurement.

The second check is the before-and-after comparison, the pivotal check of this guide. The game is opened, played for a while, and closed, and the same picture is looked at at once on the desktop. If the colours have changed, the game triggered it, and the hunt halts at this check, moving only to the profile reading that separates the layers. If the colours have not changed, the game is not the trigger, and the hunt moves to the guides for wide-gamut colour, night mode, or backlight ageing. The comparison must be made against the same picture, on the same screen, because a different picture cannot be weighed against the first one. “The colours changed” is a description from the eye, and it is only ever a change compared with before play, never an absolute judgement.

The third check is the profile reading, and it sits directly after the before-and-after comparison for a reason: the comparison proves the game triggered the change, and only the profile reading can say whether the profile itself moved or the calibration table moved underneath it. Colour Management is opened after quitting, and the default profile is read. Changed to another profile, or gone altogether, means the association layer, and the follow-up is to add the original profile back. Still the original profile, but the colours changed, means the calibration-loading layer: the gamma table was altered and not put back while the profile itself never moved. The two outcomes point opposite ways, and only this check can pull them apart.

The fourth check is the full-screen against borderless comparison. The same game is run once in exclusive full screen and once as a borderless window, and the desktop is looked at after each run. Wrong only after exclusive full screen means the full-screen mode layer: the exclusive mode lets the game write to the gamma table. Wrong after both modes means the mode is not the layer, and the hunt returns to the game, the driver, and the tools. This check comes after the profile reading on purpose, because a mode result on its own cannot say whether the profile or the calibration moved — it can only narrow down where to look once those two have been separated.

The fifth check is the high-dynamic-range and night-mode comparison. After quitting, the Windows display settings are opened and the high-dynamic-range switch is read, and the night-mode schedule is read alongside it. Switch still on means the high-dynamic-range leftover layer: the desktop is still sitting inside the high-dynamic-range container. Night-mode scheduled time matching the moment of quitting means a coincidence in time, not the game: the colours went warm because the schedule arrived, not because the game changed them. The two readings are taken from the settings, not from the look of the colours, because a washed-out look on its own cannot tell a high-dynamic-range leftover apart from a gamma table left behind, and a warm look on its own cannot tell night mode apart from a tint left behind.

The sixth check is the one-at-a-time program test. Game Bar, the monitor maker’s tool, and the calibration software are each switched off singly, and after each one the game is opened and closed again. The program that, once off, stops the trouble from coming back is the one at fault. The next step is to leave that program off or switch its automatic mode to manual as a test. If no single program stops it, the game’s own exit routine is the answer; the next step is to note which game and which display mode it was. The discipline is one variable at a time: switching several programs off together and then seeing the trouble stop settles nothing, because the test cannot say which program mattered.

The seventh check is where testing stops. The profile is gone, applying it again does not bring the colours back, and a broken picture, a black screen, or a crash has turned up next to the colour change. At that point, stop testing, and take the records in for in-store inspection: the before-and-after comparison, the profile name, the high-dynamic-range switch, and the list of checks already tried. Do not delete and install profiles one by one, and do not download a profile for an unknown screen from the internet. A profile written for another screen can make the colours worse, and trying profiles at random adds changes and no insight.

Six Misleading Ideas to Set Aside

Misleading idea one: colours changed means the screen is broken. This is wrong. Most often the shape is a gamma table left behind by a game, or a profile association problem, and the screen itself is unharmed. A screen showing wrong colours because a game left its gamma behind is a screen obeying the table it was given. Run the before-and-after comparison first, and read the profile, before deciding any hardware is at fault. A screen must never be suspected first: the game, the profile, the display mode, and the tools must all be excluded before the screen itself is blamed.

Misleading idea two: recalibrating the screen will fix it. This is wrong. Until the actor that changed the colours is named, a fresh calibration will be undone the next time the game is opened: the game, or the tool in the background, will change the table again. Calibration is the last step, not the first. The before-and-after comparison, the profile reading, and the one-at-a-time program test must run first, and only once the actor is named does a fresh calibration have a settled table to work on.

Misleading idea three: the profile is still there, so nothing is wrong. This is wrong. Choosing a profile does not equal loading its calibration. The profile is a file that states how the screen turns numbers into light; the calibration is a correction table loaded into the graphics card, and it is the table that genuinely alters the output. The profile can sit exactly where it was, untouched, while the gamma table has been altered and not put back. A reader who sees the profile still there and concludes the profile has nothing to do with it has mistaken one thing for the other. The profile and the calibration must be checked as two separate things.

Misleading idea four: it must be a driver problem. This is wrong. The game, Game Bar, the monitor maker's tool, and high dynamic range can all be the actor, and the driver is only one of the layers. Changing the driver before the before-and-after comparison has run clears away the evidence the test is built on — the baseline colours, the profile name, and the settings exactly as they were. The comparison comes first. Changing the driver before the diagnosis is done wrecks the test, and changing it once the diagnosis is done is a separate decision with checks of its own.

Misleading idea five: the driver reinstall comes first. This is wrong. The baseline goes down first, then the before-and-after comparison, then the profile reading. A reinstall attempted before the layers have been pulled apart erases the material the checks depend on — the profile exactly as it stood, the state of the calibration table, and the settings as they were. The same point holds for a state that will not change after gaming: a reinstall ahead of the diagnosis destroys the test itself.

Misleading idea six: only calibration software can change the colours. This is wrong. A game, Game Bar, a monitor maker's tool, and night mode can all change the colours, and the calibration software is only one of the actors. A reader who suspects only the calibration software will miss the game that left its gamma behind, the tool that reset the profile in the background, and the schedule that arrived at the same moment. The one-at-a-time program test exists for exactly this reason: every suspect program is switched off singly, and the one that stops the trouble names the layer.

Results Table: What Each Outcome Means

The table below collects the seven checks into one place. Every row opens from something visible — a colour shift, a profile read, a switch left on — and finishes with the layer that outcome points at. It comes after the seven checks as a second look at the diagnosis, not as a replacement for them.

What was observedLayer indicatedNext move
Colours changed only after quitting the gameGame triggered itRead the profile to separate the association layer from the calibration layer
Colours already changed before the game was openedNot the gameMove to the guides for wide-gamut colour, night mode, or backlight ageing
Profile still the original one, but colours changedFirst layer: game gamma not restored layerApply the profile again as a test; note the game and the display mode
Profile changed to another one, or goneSecond layer: profile association layerAdd the original profile back and set it as the default
Wrong only after an exclusive runThird layer: full-screen mode layerRecord which mode went wrong and which stayed normal
Wrong after both display modesNot the display modeReturn to the game, driver, and tool layers
High-dynamic-range switch still on after quittingFourth layer: high-dynamic-range leftover layerRead the switch, not the look of the colours
Night-mode scheduled time matches the moment of quittingFourth layer: coincidence in timeThe schedule arrived, not the game
One program switched off stops the troubleFifth layer: driver and monitor-tool overwrite layerLeave that program off, or switch its automatic mode to manual as a test
Profile gone, applying it again does not bring the colours back, and a broken picture, a black screen, or a crash has turned upOutside this guideStop testing and take the records in for inspection

The final two rows are worth a careful read, because they set out the limits of each test. A before-and-after comparison that shows a change proves that the game triggered it — but it does not prove that the game’s gamma is the only cause, because high dynamic range, night mode, and a monitor tool can act at the same moment. The profile reading, the switch reading, and the one-at-a-time program test must all still run before the layers can be told apart, and this piece says that limit outright. A profile still sitting in place proves nothing about the colours at all: it proves only that the profile file was not swapped, and the calibration table underneath it can still have been altered and left unrestored.

Where the Sources Part Ways

The published sources behind this guide do not speak with one voice, and the places where they part are set out plainly here instead of being smoothed over. First is the question of which cause appears most often. One published guide ranks Game Bar first. A second ranks the split between a profile gone and a calibration not loaded first. A third ranks the monitor maker’s automatic mode first. The three are cutting through the same trouble from unlike angles, not ranking one list three ways, and no settled view comes out of them. None of them is ranked first here. The guide sets out five layers, each one standing on its own evidence, and never picks one out as the most frequent cause.

The second split is about whether Game Bar is the culprit. The published account of Game Bar interfering with calibration comes from the Creators Update era, and Game Bar today is a different version. This piece does not carry the old account forward as a conclusion about the current version. The test to run is to switch Game Bar off and play again. A reader who takes the old account as a present-day fact will blame a program that may have nothing to do with the shape in front of them.

The third split is about a profile that will not stay put. One published discussion, in a Microsoft question-and-answer thread, describes a custom profile failing to stay in place after a restart or after waking from sleep, with causes including a driver reset, high dynamic range, and automatic colour management. That discussion is about restarting and waking, not about quitting a game, and this piece does not carry its account across to a game exit. The two triggers are unlike, and the checks for one cannot be used as the checks for the other. The thread is drawn on here only for the mechanism — that a driver reset or high dynamic range can displace a profile — and not for the trigger.

The fourth split is about whether high dynamic range makes a profile useless. Two published guides state that with high dynamic range on, an ICC profile can become ineffective, and desktop colours look washed out and grey. Both are describing colours while high dynamic range is on, not colours after quitting with high dynamic range left on. This piece keeps the two apart: colours wrong while high dynamic range is on are covered by the high-dynamic-range guide, and colours wrong after quitting because high dynamic range was left on are covered here. The switch reading after quitting is the test that separates them, and the look of the colours on its own cannot.

The last split is over whether reapplying the profile counts as a repair or a test. Published guides on profiles that keep resetting treat putting the profile back on as a step that brings the colours back. The position here is different: putting the profile back on is a test. It shows the trouble sits in the calibration table, because reloading the table restores the colours, and it does not fix the actor that changed the table. The split matters: a reader who treats the reapplication as a repair stops there, at the very point the layer has just been identified, without ever looking to see which game, tool, or schedule changed the table in the first place.

Patterns That Belong to Other Guides

The table below collects the shapes that resemble this one but belong to other guides. Each row pairs a shape with its proper guide, and the reason it sits outside this piece. The table does diagnosis work of its own: when a row here fits the shape in front of the reader better than the opening section’s shape does, the reader should switch guides.

ShapeThe guide that covers itWhy it belongs to another guide
Wide-gamut screen shows standard colours too strongly while viewing#237, wide-gamut colourThe colours are already too strong while viewing, not changed after quitting
Two screens show the same picture in unlike colours#201, two-screen calibrationThe premise is a calibration difference between screens, not a game trigger
Colours wrong while high dynamic range is on#139, high dynamic rangeThe premise here is colours wrong after quitting, not while high dynamic range is on
Picture suddenly turns black and white, colours gone#364, colour filter and driver colour settingsThe premise here is a colour shift after quitting, not a sudden loss of colour
Video edit preview normal, finished file colours unlike#499, colour spaceThe premise is a colour-space difference in output, not a game exit
Resolution or refresh rate lower after waking from sleep#322, wake re-handshakeThe trigger is waking, not quitting a game
Screen looks yellow or warm all the time#233, night mode and low blue lightThe premise is a standing warm cast, not a change after quitting
Screen looks grey and colours pale#250, full against limited colour rangeThe premise is a range mismatch, not a game exit
Screen grows yellower the longer it is used#410, backlight ageingThe premise is a gradual shift over time, not a step after quitting
Screen flickers at low brightness#387, pulse-width dimmingThe premise is flicker, not a colour setting
Colours inside the game itself are already wrongNot this guideThe premise here is colours normal inside the game and wrong only after quitting
Profile lost after a restart or a wake from sleepRestart and wake guideThe trigger is restarting or waking, not quitting a game

Two rows in that table call for a second look: readers mistake them for this shape more often than any others. The first is the wide-gamut guide. A wide-gamut screen showing standard colours too strongly is already wrong while the content is on screen, and its checks turn on the colour space and the profile. This shape is a change that shows up only after quitting, and its checks turn on the game, the gamma table, and the profile association. The two shapes share the word “colour” and little else, and their checks run in different directions. Next comes the guide on waking from sleep. Waking to a smaller picture or a slower refresh is a wake fault, not a game fault. The game in that shape is only there by chance, so running the before-and-after comparison there settles nothing, because the fault tracks waking rather than the game. Mixing the two up puts the reader on the wrong set of checks.

When to Stop Testing

Stop once no layer has come out of the seven checks. A picture pinned to one wallpaper, compared before play and after quitting, read in Colour Management, tried in both display modes, checked against the high-dynamic-range switch and the night-mode schedule, and tried with the suspect programs switched off one at a time — and still with no clear answer — has given up everything the home checks can give. Pushing on from there is guessing, and guessing spent on profiles that were never at fault only muddies the diagnosis. Restore the test options before stopping, leaving the machine settled.

Stop as well when the profile is gone, applying it again does not bring the colours back, and a corrupted picture, a black screen, or a crash has shown up beside the colour change. That pairing settles the question by itself: the home checks have cleared the game layer, the profile layer, the display-mode layer, and the tool layer, and what remains lies beyond what the home checks can tell apart. Stop too after changing one setting with no shift, and changing a second setting with no shift either. Two tried changes and an outcome that has not moved settle the question by themselves: those settings do not run the layer. Going on to a third, fourth, and fifth change adds nothing. An inspection after that has to start by undoing the whole list, back to the start.

Stop at once, and run no more checks, when a fresh symptom shows up mid-testing, with the screen going black, the picture breaking up, the machine restarting, or the screen reporting no signal. Those symptoms point to a different family of guides, and going on with colour checks while they are showing can turn a check into repair work. Note down the new symptom, when it showed up, and which check was running, and take that note with the rest. That note goes with the evidence, and it counts for more than every list of changed settings.

Keep the baseline notes and the screen photographs once testing is done. The default profile name, the high-dynamic-range switch, the night-mode switch, the before-and-after comparison of the same picture, and the photographs of the screen, taken before and after every check, together count for more than any one of them on its own. Those notes are what every inspection starts from, and that is why the opening check asks for the picture and the settings to be pinned down and photographed before any testing begins.

What to Bring for Inspection

Bring the seven checks’ records with you: the pinned picture and how it looked before play and after quitting; the default profile name before and after; the result of the before-and-after comparison; which display modes stayed normal and which went wrong; the high-dynamic-range switch reading after quitting; the night-mode schedule reading; and which suspect program, if any, stopped the trouble when it was switched off.

Bring the photographs taken in the opening check: how the colours shifted is evidence in its own right, and a description remembered afterwards is no record at all. Bring the game name, the display mode in use, and the screen model, because that trio counts as evidence too. Leave the machine with the settings as they stand now, every test setting put back, and drop any record of settings that were changed and then restored. What matters is the settings as they stand now, plus what the seven checks gave back. The history of changes is worth something only where a change altered the colours; such a change names the layer.

For inspection, HK GPU Fix is at Dali Plaza 1209, Shek Mun. The phone number is 2186 8010 and WhatsApp is 6558 6806. The published inspection fee range is HK$400 to HK$2,800, and no price is estimated before inspection. There is no Mong Kok shop and no on-site service.

Frequently Asked Questions

My desktop colours changed after I quit a game. Is my monitor broken?

Almost certainly not. Most often the shape is a gamma table left behind by the game, or a profile association problem, and the screen itself is unharmed. The test is the before-and-after comparison: look at the same wallpaper or photograph before the game is opened and again immediately after it is closed. If the colours changed only after quitting, the game triggered it. Then open Colour Management and read the default profile: still the original one with colours changed points at the calibration table, while a profile changed to another one points at the association layer. A screen showing wrong colours because a game left its gamma behind is a screen obeying the table it was given, not a screen that has failed.

My colour profile is still listed in Colour Management. Does that mean nothing is wrong?

No. Choosing a profile does not equal loading its calibration. The profile is a file that states how the screen turns digital numbers into light; the calibration is a correction table loaded into the graphics card, and it is the table that genuinely alters the output. The profile can sit exactly where it was, untouched, while the gamma table has been altered by a game and not put back on exit. A reader who sees the profile still there and concludes the profile has nothing to do with it has mistaken one thing for the other. The profile and the calibration must be checked as two separate things, and the checks in this guide do exactly that.

Should I just recalibrate my screen to fix it?

No — not before the actor is named. Until the program that changed the colours is found, a fresh calibration will be undone the next time the game is opened: the game, or the tool in the background, will change the table again. Calibration is the last step, not the first. Run the before-and-after comparison first, read the profile, run the display-mode comparison, and switch the suspect programs off one at a time. Only once the actor is named does a fresh calibration have a settled table to work on. Applying the profile again is a test that shows the trouble sits in the calibration table; it is not a repair.

It only happens after exclusive full screen, not after a borderless window. Why?

Because in exclusive full screen the game can write to the gamma table for its own picture, while borderless windowed play keeps the desktop composition in charge of the picture. A game that changes the table for its own picture must put it back when it closes, and some games do not. In borderless mode the game cannot change the table in the same way, so the colours stay normal after quitting. The mode comparison is what names the layer; playing borderless sidesteps the shape without fixing the game’s exit routine.

High dynamic range was on in the game. Is that the cause?

It can be, but only the switch reading after quitting can say so. Open the Windows display settings after quitting and read the high-dynamic-range switch: still on points at the leftover layer, with the desktop still sitting inside the high-dynamic-range container, and standard content looking washed out or grey. Never switched on in the game at all points away from high dynamic range. Two shapes must be kept apart: colours wrong while high dynamic range is on, which belongs to the high-dynamic-range guide, and colours wrong after quitting because high dynamic range was left on, which belongs here. The switch is the evidence, not the look of the colours.

Night mode came on at the same time I quit the game. Did the game cause it?

Most likely not — the two arrived together by coincidence in time. Night mode changes the colour temperature on a schedule, making colours look warmer, and its scheduled time can match the moment a game is closed. Read the night-mode schedule and compare it with the moment of quitting: a scheduled time matching that moment points at a coincidence, not at the game. A coincidence in time is not a cause: the two arriving together proves nothing. Switch night mode off as a test, or change its schedule, and play again; if the colours stay normal, the schedule was the actor.

Could Game Bar or my monitor maker's tool be changing the colours?

Yes, both are suspect programs, and both must be tested one at a time. Game Bar in older versions interfered with picture calibration and ICC profiles, though that account comes from the Creators Update era and Game Bar today is a different version, so it can only be tested, not assumed. A monitor maker's tool in automatic mode can keep resetting the external screen's profile back to the one the tool itself chooses, and published guides state that switching the tool to manual mode stopped the resetting in the cases they describe. Switch each program off singly, open and close the game again after each one, and note which one, once off, stops the trouble from coming back. Switching them all off at once finds nothing.

What is my monitor maker’s tool’s Auto Mode?

Auto Mode is a setting in some monitor makers’ tools that keeps putting the external screen’s profile back to the one the tool itself picks. That is why a profile can be set correctly one moment and changed back after a single game: the game did not change it — the tool in the background did. The test is to switch the tool to manual mode and try again: if the resetting stops, the tool layer is the actor. Published guides state that switching the tool to manual mode stopped the resetting in the cases they describe.

The colours come back after I restart the machine. Does that mean it is fixed?

No. Restarting the machine reloads the correction table, and the colours return because the table was restored — not because the actor was found. The next time the game is opened and closed, the same game can change the table again and leave it unrestored. A restart proves the trouble sits in the calibration table, because reloading the table restores the colours, and it does not fix the game's exit routine, the tool in the background, or the leftover high-dynamic-range state. Run the before-and-after comparison and the profile reading, and name the actor, before calling the shape fixed.

When do I stop testing and book an inspection?

Stop when all seven checks are done, the profile is gone, applying it again does not bring the colours back, and a corrupted picture, a black screen, or a crash has shown up beside the colour change. Bring the seven checks’ records: the pinned picture and how it looked before play and after quitting, the default profile name before and after, the before-and-after comparison result, which display modes stayed normal and which went wrong, the high-dynamic-range switch reading, the night-mode schedule reading, and which suspect program if any stopped the trouble — with the opening check’s photographs, the game name, the display mode in use, and the screen model. Those records, plus the settings as they now stand, are what an inspection builds on.

Sources

The list below gives the sources this article drew on. They are used for the mechanism and the test method, not as things to follow, and figures from them are quoted only as their claims, wherever the text marks them as such. No figure from any source is repeated here as a finding of this site.

  1. drivethelife.com, "Fix: Screen Calibration Resetting after Windows 10 Creators Update" — https://www.drivethelife.com/screen-calibration-resetting-after-windows-10-creators-update/ — after entering a full-screen game the picture calibration resets, the brightness changes, and the profile stops taking effect; that site states that Game Bar interfered with calibration and ICC profiles, which is a Creators Update era account and is not carried forward here as a conclusion about the current version; night mode and driver updates are listed by that site as possible causes.
  2. windowsforum.com, "Windows 11/10 ICC Color Profile Keeps Resetting? Fix Display Calibration, HDR and Driver Overrides" — https://windowsforum.com/tutorials/windows-11-10-icc-color-profile-keeps-resetting-fix-display-calibration-hdr-and-driver-overrides.448948/latest — a profile still selected with colours changed, and a profile gone, are two unlike problems and must be checked separately; the profile must first be confirmed as genuinely gone before the calibration loading and driver overwrite are checked; with high dynamic range on, an ICC profile can become ineffective.
  3. us.ktcplay.com, "Monitor Calibration Profile Not Loading? Here's the Fix" — https://us.ktcplay.com/blogs/support-tips/monitor-calibration-profile-not-loading-fix — a profile disappearing after a restart or after waking from sleep is most often the platform replacing it, not the screen failing; a colour change appearing only when switching windows, opening a game, or opening a monitor tool points towards a software conflict; a monitor tool's automatic mode kept resetting an external screen's profile, and switching the tool to manual mode stopped it.
  4. learn.microsoft.com, "windows 11 color profile keeps changing", Microsoft Q&A — https://learn.microsoft.com/en-ca/answers/questions/5720286/windows-11-color-profile-keeps-changing — a custom ICC profile can fail to stay in place after a restart or after waking, with causes including a driver reset, high dynamic range, and automatic colour management; that discussion is about restarting and waking, not about quitting a game, and its account is not carried across here to a game exit.
  5. deadpixeltests.com, "How to Fix Monitor Color Profiles (ICC) in Windows & Mac" — https://deadpixeltests.com/blog/how-to-fix-monitor-color-profiles/ — an ICC profile is the standard file stating how a screen turns digital red, green, and blue numbers into actual light output; a profile that does not fit, is damaged, or is overwritten by a driver or by high dynamic range harms colour accuracy; Windows 11 with high dynamic range on forces desktop programs into the high-dynamic-range container, and on a screen whose peak brightness is not high, desktop colours look washed out and grey.

Desktop colours that were normal before play and wrong after quitting are not a verdict on the screen. Take one picture, look at it before the game and again the moment the game is closed, read the profile in Colour Management, try the two display modes, read the high-dynamic-range switch and the night-mode schedule, and switch the suspect programs off one at a time. Those checks show whether the game’s exit routine, the profile association, the display mode, a leftover state, or a tool in the background changed the colours — and they show it without touching the hardware.

Related articles

Related services

If you need Chip-level GPU repair, see these service pages: