Game FPS Halved After Sleep Wake, Restarting the Game Fixes It? Resume Power State, Display Mode and Game State Diagnosis Guide

2026-11-18 | GPU Repair | 43 minutes | HK GPU Fix Editorial Team

Game FPS Halved After Sleep Wake, Restarting the Game Fixes It? Resume Power State, Display Mode and Game State Diagnosis Guide

Key Summary

After waking from sleep, a game runs at about half its usual frame rate, and restarting only the game sometimes restores it. This guide uses three comparisons — restarting the game, checking the clock speed, and restarting the computer — to separate a game state fault from a driver power state fault, so the right layer is tested first.

The Shape of the Problem: Half Speed After Waking, Full Speed Before

This guide covers one specific shape, and the shape itself is the whole diagnosis. The computer goes to sleep. It wakes up. A game is launched, or a game that was already running is brought back to the foreground, and the frame rate is about half of what it was before the computer slept. For example, a game that ran at 100 frames per second before sleep now runs at 50. A game that held a steady 60 now holds a steady 30. The picture is intact — there is no tearing that was not there before, no coloured block, no black screen, no crash — only the speed has dropped, and it has dropped by roughly half.

Three details define the shape, and all three have to be present. First, the slowness appears only after waking from sleep. Before sleep, the same game in the same scene ran at its normal speed. Second, the drop is large and stable: the frame rate does not bounce between normal and slow, it sits at roughly half speed and stays there. Third, the rest of the computer behaves normally enough that the game can still be played — the desktop responds, other programs open, and nothing has visibly failed. A game that slows down gradually over a long session, a game that stutters suddenly in the middle of play, or a computer that wakes to a black screen are all different shapes, and they belong with different guides.

That combination is what makes this symptom different from its neighbours. One neighbouring shape is a computer that feels slow in general after waking — games stutter, the desktop feels heavy, and everything returns to normal after a restart. That shape is covered in the guide numbered 284, and it uses a restart comparison as its main test. The guide numbered 284 never treats restarting only the game as a test at all: the words for restarting the game do not appear in it once. This guide adds exactly that test, because the test splits the problem in two at almost no cost. Another neighbouring shape is a game that gets slower and slower the longer it is played, which is covered in the guide numbered 356 — that shape is triggered by playing, not by waking. A third neighbour is a game that suddenly stutters or drops frames in the middle of play, covered in the guide numbered 136, which works through five paths — cooling, power supply, driver, settings, and hardware — and never mentions sleep or waking at all. A fourth neighbour is a laptop game locked at 30 frames per second until the charger is plugged in, covered in the guide numbered 362, which is a battery mode problem, not a wake problem. A fifth neighbour is a game whose frame rate number looks fine but whose play still feels jerky frame by frame, covered in the guide numbered 363, which is a frame time problem, not a halved frame rate. Each of those guides has its own shape, and this guide starts from a different one: normal before sleep, halved after waking, and the question of whether restarting only the game restores it.

The most common first conclusion is that the graphics card must be failing. That conclusion jumps too far. A card that is failing does not usually fail only after waking and then run perfectly before the next sleep. What this shape points at is a state that was not restored properly: either the game itself did not recover its display mode and device context when the system woke, or the driver and the power state did not return the card to its normal operating level. Those are two different layers, and they are separated by one cheap test that takes less than a minute. The rest of this guide is about running that test properly, reading its result honestly, and then following the branch the result points to, instead of replacing a card that was never the problem.

One practical point has to be settled before anything else. A baseline must be written down before any test is run. The baseline is the frame rate of the same game, in the same scene, before the computer slept, together with the display mode the game was using — exclusive fullscreen or borderless window — and the power plan that was active. Without that baseline, "about half" is a guess, and a guess cannot separate a halved frame rate from a game that was always slow in that scene. The baseline should be a number, read from the game's own counter or from a monitoring tool, not a feeling remembered afterwards. The same applies to the result after waking: the frame rate after waking, in the same scene, read the same way. Two numbers, taken the same way, are the evidence. Everything in this guide builds on them.

How Waking Restores a System in Stages

Waking from sleep is not a single action. It is a sequence of restorations, and any one of them can go wrong while the others succeed. Windows has to restore the kernel state. The graphics driver has to restore the state of the graphics card. The card has to move from a low-power state back to its normal clock speed. The game has to restore its own display device context and display mode. A fault in any one link of that chain can leave the frame rate at about half, because the game ends up running through a path that was never meant to carry full-speed play.

The driver is a kernel-mode component, and its behaviour after waking is the first mechanism to understand. After waking, the driver or the power state can remain stuck at a low-performance level, so the card never returns to its normal operating level. The card is not broken and it is not overheating; it simply has not been woken up properly. Two user reports on the Tom's Hardware forum describe exactly this. In one, an AMD driver left the core clock soft-locked at 150 MHz after waking, and the problem only went away when an older driver version was installed. In the other, the core clock was locked at 658 MHz with the temperature at only 50 degrees Celsius, and the clock only jumped back to 1304 MHz after a monitoring tool was opened and a stress test was run. Those numbers are forum user reports, not measurements made for this guide, but they establish the mechanism clearly: a clock can be locked low while the temperature is completely normal. That is the second counter-intuitive core of this guide, and it has to be stated plainly — a normal temperature does not mean the card is fine, and it does not mean the card is running at full speed. Temperature only shows how hot the card is. It says nothing about what clock speed the card is actually running at. A card that has never been woken up properly runs cool precisely because it is doing so little work.

The game has a state of its own, and that is the second mechanism. A game running in exclusive fullscreen binds itself to the display output at the moment it starts. When the system wakes, the display is re-enumerated and the display mode may change, so the device context the game is holding may no longer match the output it is bound to. The game does not always rebuild that context on its own. It keeps running through a wrong, low-performance path instead. Restarting the game gives the game a clean start: it binds the output again, rebuilds its device context, and the frame rate returns. That is why restarting only the game sometimes restores full speed without any restart of the computer. The fault was never below the game; it was in the game layer itself.

Fast Startup makes the situation more complicated, and it is the third mechanism. With Fast Startup switched on, shutting the computer down is not a clean start from scratch. It is a partial hibernation: the kernel memory image, including the kernel-mode drivers that are already loaded, is written to the hibernation file and loaded back at the next switch-on. If the driver state has already gone bad after waking, Fast Startup carries that bad state faithfully into the next session. A restart is different. A restart performs a complete boot cycle, Fast Startup does not apply to a restart, and the whole graphics stack is built again from the beginning. That is why "shut down and switch on again" and "restart" are not the same thing, and why this guide insists on a real restart for its restart comparison. The distinction is explained in the guide numbered 207, which covers the Fast Startup mechanism in full; this guide uses only the one concept it needs, that a restart is not the same as shutting down and switching on again.

There is also a fourth mechanism that belongs to laptops. On a laptop with two graphics processors, the game may run on the integrated graphics after waking because the discrete card was never called on. The frame rate halves because the game is running on the weaker processor, not because the discrete card is slow. The test for that layer is to look at which processor the game is actually using, and whether the discrete card's usage is sitting at zero. That check costs nothing and it separates a switching fault from a clock fault in one look.

A fifth mechanism is background load. When a system wakes, many background tasks resume at the same time and compete for resources, which can make a game look slower than it really is. That is why the background layer has to be excluded early: a game that is slow because the system is busy after waking is not a graphics card problem at all, and every test run after that point would be read against the wrong baseline. The guide numbered 284 already states this order — exclude background programs before concluding that the card is at fault — and this guide keeps the same order.

The Deciding Forks: Three Comparisons That Narrow the Field

The whole guide turns on three forks. Each fork is a comparison, and each comparison removes a whole set of possibilities. They should be done in order, before any setting is changed, because a setting changed before a fork is tested destroys the evidence the fork would have given.

Fork One: Does Restarting Only the Game Restore the Frame Rate?

The first fork is the game restart comparison, and it is the cheapest test in the guide. Change nothing else. Close the game completely, open the same game again, load the same scene, and read the frame rate. If the frame rate returns to normal at once, the fault belongs in the game and application state layer: the game did not recover its display mode or device context when the system woke, and a clean start fixed the game's own state. If the frame rate is still halved, the fault belongs below the game, in the driver or power state layer, and restarting the game cannot reach it. That is the decisive branch of this guide, and it is why this test comes first: one comparison separates the game layer from the driver layer, and it decides which road the rest of the diagnosis follows.

An honest statement has to be made here, because it is the point on which this guide is most often misread. Outside sources say that restarting the game is usually unhelpful and that a restart of the computer is what actually helps. A guide on w3tutorials puts it plainly: restarting the game or lowering the graphics settings is almost no help, and restarting Windows is what fully restores performance. That is the common case, and this guide does not dispute it. In the common case the fault is in the driver or power state layer, which restarting the game cannot touch. This guide does not claim that restarting the game is a fix. It claims that restarting the game is a decisive test. Run it and the frame rate returns, and the fault is locked into the game layer. Run it and the frame rate stays halved, and that is not a failure — it is evidence, proving the fault is in the driver or power state layer and that the road described in the guide numbered 284 is the right one to follow. Presenting the game restart as a universal fix would be wrong. Presenting it as useless would be equally wrong, because it would throw away the one test that can separate the two layers in under a minute. Both readings have to be kept in the same sentence: usually unhelpful as a fix, decisive as a test.

Fork Two: Is the Clock Speed Locked Low While the Temperature Is Normal?

The second fork is the clock speed comparison. Open GPU-Z or HWiNFO and look at the core clock speed, the power draw, and the temperature while the game is running. A core clock locked at a low level — a few hundred megahertz, for example — while the temperature sits at a normal value points at the driver power state layer: the card has not been woken up, and it is not a question of the card lacking power. A normal clock speed with the frame rate still halved points at the game and application state layer or at another layer, because the card has woken up and the fault is in how the game is using it. The numbers have to be written down. A feeling that the game "seems slow" is not a measurement, and a temperature reading on its own is not a measurement of clock speed either. The two numbers together — clock speed and temperature — are what this fork reads, and they have to be read in the same scene, at the same settings, as the baseline.

Fork Three: Is Only One Game Slow, or Are All Games Slow?

The third fork is the range comparison. Is only one game slow, or are all games slow after waking? One game slow leans toward the game itself: the fault is in that game's own state, and the game restart comparison will usually confirm it. All games slow leans toward the driver or power state layer: the fault sits below every game, and no single game can be blamed for it. This fork is done third because the first two forks are more precise, but it is still worth doing, because it catches the case where a game-specific fault is mistaken for a system-wide one, or the other way around. The three forks should be done in order: the game restart comparison first, because it is the cheapest and it decides the road; the clock speed comparison second, because its result decides whether the investigation follows the driver layer or the game layer; and the range comparison third, as a check on the first two.

ForkComparisonResult AResult B
Game restart comparisonClose the game and reopen the same game in the same scene, changing nothing elseFrame rate returns at once — game and application state layerFrame rate still halved — driver or power state layer
Clock speed comparisonCore clock, power draw, and temperature in GPU-Z or HWiNFO while the game runsClock locked low, temperature normal — driver power state layerClock normal, frame rate still halved — game layer or another layer
Range comparisonThe same test in a second game after wakingOnly one game slow — that game's own stateAll games slow — driver or power state layer

The Five Suspicion Layers

The five layers below are the whole map of this symptom. They are ordered by the cost of testing them, not by how likely they are, and the order itself is part of the diagnosis. The game restart comparison is first because it costs almost nothing and it separates the two largest branches. The clock speed comparison is second because it decides whether the driver layer or the game layer is followed next. The background layer has to be excluded early, because an unexcluded background load will corrupt every comparison that follows.

LayerWhat it isDecisive test
First layer: game and application stateThe game did not recover its display mode or device context after waking, and an exclusive fullscreen game is bound to the wrong outputGame restart comparison: close and reopen the game, frame rate returns at once means this layer, still halved means move on
Second layer: driver power stateThe driver or the power state stays at a low-performance level after waking, and the core clock is locked lowClock speed comparison: GPU-Z shows the core clock locked at a low level, even though the temperature is normal
Third layer: Windows power and Fast StartupFast Startup carries a degraded driver state into the next session, and the power plan leaves the card in a power-saving stateRestart comparison and Fast Startup comparison: a restart restores speed while shutting down and switching on again does not, pointing at this layer
Fourth layer: discrete and integrated graphics switchingOn a laptop, the game runs on the integrated graphics after waking because the discrete card was never called onDevice comparison: check which card the game is using in Task Manager or in the game's own display settings, and whether discrete card usage is zero
Fifth layer: background programs and loadMany background tasks resume at the same time when the system wakes, competing for resources and making the game look like a graphics card problemBackground comparison: close the background programs and compare the same workload before and after waking, as the guide numbered 284 already states

First Layer: Game and Application State

This is the layer the first fork tests. When a system wakes, the display is re-enumerated and the display mode may change. A game that bound itself to the display output when it started may now be holding a device context that no longer matches the output. The game does not always notice, and it does not always rebuild. It keeps running through a wrong, low-performance path. The symptom is precise: the frame rate halves after waking, and restarting only the game restores it at once. Nothing below the game needs to change, because nothing below the game is wrong. The follow-up test for this layer is the display mode comparison in step four: switching the game from exclusive fullscreen to borderless window and back, or re-selecting the display and the resolution inside the game's settings, applying the change, exiting the game normally so the settings are saved, and opening the game again to see whether it remembers. If the frame rate returns after a mode switch, the fault is confirmed in the game's display mode recovery. If no mode change makes any difference, the investigation moves to the fifth step.

Second Layer: Driver Power State

This is the layer the second fork tests. The driver or the power state stays at a low-performance level after waking, and the core clock is locked low. The card is not overheating and it is not failing; it has not been returned to its normal operating level. The symptom is precise in the opposite direction from the first layer: the frame rate stays halved after the game is restarted, and the clock speed reading shows the lock. The two forum reports quoted earlier belong to this layer: a core clock soft-locked at 150 MHz after waking, fixed only by installing an older driver version, and a core clock locked at 658 MHz at 50 degrees Celsius, released only after a monitoring tool was opened and a stress test was run. Both are user reports from a forum, and they are quoted as user reports, not as measurements made for this guide. They show the mechanism, and they show why the temperature reading must never be read on its own. A card that is locked at a low clock runs cool because it is doing so little work. Cool does not mean healthy. It means idle.

Third Layer: Windows Power and Fast Startup

This layer is about how the system starts, not about how the game runs. With Fast Startup switched on, shutting down is a partial hibernation, and the kernel memory image — including the loaded kernel-mode drivers — is written to the hibernation file and loaded back at the next switch-on. If the driver state has already gone bad after waking, Fast Startup carries that bad state into the next session faithfully. A restart is a complete boot cycle, and Fast Startup does not apply to a restart. The comparison that tests this layer is therefore a restart against a shut-down-and-switch-on: if a restart restores the frame rate while shutting down and switching on again does not, the fault points at this layer. The mechanism is explained in full in the guide numbered 207, and this guide uses only the one concept it needs. The practical consequence is simple: when a reader says "I already restarted and it did not help", the first question has to be whether they really restarted, or whether they shut the computer down and switched it on again, because the two are not the same and a comparison built on the wrong one is worthless.

Fourth Layer: Discrete and Integrated Graphics Switching

This layer belongs to laptops with two graphics processors. After waking, the game may run on the integrated graphics because the discrete card was never called on. The frame rate halves because the weaker processor is doing the work, not because the discrete card is slow. The test is to look at which card the game is actually using, in Task Manager or in the game's own display settings, and to check whether the discrete card's usage is sitting at zero while the game runs. If the discrete card is at zero, the fault is in the switching, not in the card. If the discrete card is working and its clock speed is normal, this layer is excluded and the investigation returns to the game layer or the driver layer. This check is cheap, and it should be done before any driver change is considered, because changing a driver cannot fix a game that was never sent to the right processor.

Fifth Layer: Background Programs and Load

This layer is the one that has to be excluded before any other conclusion is trusted. When a system wakes, many background tasks resume at the same time — updates, synchronisation, indexing, and scheduled work — and they compete for the processor, the memory, and the disk. A game measured while that storm is still running will look slower than it really is, and every comparison made during that window will be wrong. The test is to close the background programs, wait for the system to settle, and then compare the same workload before and after waking. The guide numbered 284 already states this order, and this guide keeps it: do not conclude that the card is at fault until the background layer has been excluded, because the direction of the whole investigation depends on it.

The Seven Diagnostic Steps

Work through these steps in order. Each step is a comparison, and each comparison produces a written result. Change only one thing at a time. A step that changes two things at once produces a result that cannot be read, because no one can tell which change did the work.

StepWhat to doWhat the result means
Step 1: Record the baselineWrite down the frame rate before sleep and after waking, in the same game and scene, plus the game name, display mode, and power planWithout a baseline, "about half" is a guess, and the rest of the diagnosis has no evidence
Step 2: Clock speed comparisonOpen GPU-Z or HWiNFO and read the core clock, power draw, and temperature while the game runsClock locked low with normal temperature points at the driver power state layer; normal clock points at the game layer or another layer
Step 3: Game restart comparisonChange nothing else; close the game and reopen the same game in the same sceneFrame rate returns at once means the game layer; still halved means the driver or power state layer
Step 4: Display mode comparisonOnly if step 3 points at the game layer: switch between exclusive fullscreen and borderless window, or re-select the display and resolutionFrame rate returns after a mode switch points at the game's display mode recovery; no change means move to step 5
Step 5: Driver reset hotkey comparisonPress Win + Ctrl + Shift + B; the screen goes black for a moment and returnsFrame rate returns at once points at the driver state layer; no change means move to step 6
Step 6: Restart comparisonDo a real restart, not a shut-down-and-switch-on, then open the same game and compareA restart restores speed points at the driver or power state layer; still slow means the fault is not in the wake layer
Step 7: Decide and stopWrite down every comparison result and decide which layer the evidence points atIf the seven steps cannot separate the layers, or if the slowdown comes with a blue screen, a black screen, a driver crash message, or a burning smell, stop and bring the records for inspection

Step 1: Record the Baseline

The first step is to describe the fault precisely. How far has the frame rate fallen — roughly half, or down to a dozen frames per second? Is only one game affected, or are all games affected? What was the frame rate of the same game, in the same scene, before the computer slept? Write down the baseline: the frame rate before sleep, the frame rate after waking, the game name, the display mode — exclusive fullscreen or borderless window — and the power plan. A number read from the game's own counter or from a monitoring tool is evidence. A feeling remembered afterwards is not. This step takes a few minutes and it protects every step that follows, because a diagnosis built on a guessed baseline is a guess all the way down.

Step 2: The Clock Speed Comparison

Open GPU-Z or HWiNFO and look at the core clock speed, the power draw, and the temperature while the game is running. A core clock locked at a low level with a normal temperature points at the driver power state layer: the card has not been woken up, and the problem is not that the card lacks power. A normal clock speed with the frame rate still halved points at the game and application state layer or at another layer, because the card has woken up and the fault is in how the game is using it. Write the numbers down. Do not report that the game "seems slow" without the clock speed next to it, and do not report that the temperature is normal as if that settled the question. The temperature only shows how hot the card is. The clock speed shows how fast it is actually running. The two numbers answer different questions, and this step needs both.

Step 3: The Game Restart Comparison — The Decisive Test

Change nothing else. Close the game completely, open the same game again, load the same scene, and read the frame rate. If the frame rate returns to normal at once, the fault is locked into the game layer, and the investigation moves to step four. If the frame rate is still halved, the fault is below the game, and the investigation moves to step five. This is the decisive test of the whole guide, and it has to be run cleanly: the same game, the same scene, the same settings, nothing else changed. Changing a setting at the same time as restarting the game destroys the comparison, because no one can then tell whether the restart or the setting did the work. The result also has to be read honestly. A frame rate that returns after restarting the game is not proof that the game is fixed forever — it is proof that the fault belongs in the game layer, and that the next question is which part of the game's state failed to recover. A frame rate that stays halved is not a failure of the test — it is evidence that the fault belongs below the game, and it is the evidence that sends the investigation toward the driver and power state layer.

Step 4: The Display Mode Comparison

This step is only done when step three points at the game layer. Switch the game from exclusive fullscreen to borderless window and back again, or re-select the display and the resolution inside the game's settings, apply the change, exit the game normally so the settings are saved, and open the game again to see whether it remembers. If the frame rate returns after a mode switch, the fault points at the game's display mode recovery: the game was holding a display mode that no longer matched the output after waking, and the switch forced it to rebuild. If no mode change makes any difference, move to step five. This step is deliberately narrow. It tests only the game's display mode handling, and it does not change the driver, the power plan, or anything else.

Step 5: The Driver Reset Hotkey Comparison

Press Win + Ctrl + Shift + B. This combination restarts the display driver: the screen goes black for a moment and then returns, and there is no need to restart the whole computer. If the frame rate returns at once after the hotkey, the result points at the driver state layer. If nothing changes, move to step six. The hotkey is a comparison, not a repair. It clears the driver state for one test, and it must not be written up as "fixed". A driver state that has to be cleared by hand after every wake is a driver state that has not been restored properly, and the question of why it fails to restore is still open. The hotkey tells the reader which layer the fault belongs to. It does not tell the reader that the fault is gone.

Step 6: The Restart Comparison

Do a real restart — not a shut-down-and-switch-on, because Fast Startup makes shutting down a partial hibernation rather than a clean start — and then open the same game and compare the frame rate. If a restart restores the speed, the result points at the driver or power state layer, and the evidence is already enough: there is no need to keep guessing. If a restart does not restore the speed either, the fault is not in the wake layer at all, and the investigation should follow the road described in the guide numbered 136 — the five-path check through cooling, power supply, driver, settings, and hardware. The restart comparison is the most expensive test in the guide, which is why it comes sixth rather than first. Running it first would destroy the evidence that the cheaper tests would have given, and it would leave the reader knowing only that a restart helps, without knowing which layer the fault belongs to.

Step 7: Decide and Stop

Once a layer has been found, write down every comparison result: the game restart result, the clock speed numbers, the driver reset result, and the restart result. A written record is what makes the next step possible, whether that step is a settings change, a driver version comparison, or an inspection. If all seven steps have been run and the layers still cannot be separated, or if the frame rate drop comes with a blue screen, a black screen, a driver crash message, or a burning smell, stop testing and bring the full record for inspection. Continuing to change settings after the diagnosis has failed to separate the layers only adds new variables, and a burning smell is a stop condition in itself, not a symptom to test through.

Six Misconceptions That Mislead This Diagnosis

1. "The frame rate drops after waking, so the card must be broken."

Wrong. Slowness after waking is usually a state recovery problem — the driver stays at a low-performance state, or the game does not recover its display mode. A state problem and a hardware fault are two different things. Until the game restart comparison and the restart comparison have both been run, the hardware must not be condemned. A card that runs perfectly before sleep and halves only after waking is describing a restoration failure, not a dying card.

2. "Restarting the game is useless, so skip it and restart the computer."

Wrong. Restarting the game is not a repair suggestion; it is a decisive test. Run it and the frame rate returns, and the fault is locked into the game layer. Run it and the frame rate stays halved, and that is itself evidence, proving the fault is in the driver or power state layer. Skipping this step makes it impossible to separate the two sides. The outside sources are right that restarting the game is usually unhelpful as a fix — in the common case the fault is below the game. But "usually unhelpful as a fix" and "useless as a test" are different claims, and this guide keeps them apart.

3. "Shutting down and switching on again is the same as restarting."

Wrong. With Fast Startup switched on, shutting down is only a partial hibernation, and the kernel driver state is carried back into the next session. Only a restart performs a complete boot cycle. A guide on w3tutorials states this plainly: the Fast Startup setting does not apply to a restart. When a comparison is run, it has to use a real restart, or the result of the comparison is itself unreliable.

4. "The temperature is normal, so there is no throttling."

Wrong. A forum report on Tom's Hardware describes a core clock locked at 658 MHz with the temperature at only 50 degrees Celsius. A clock can be locked low while the temperature is completely normal, because the card was never woken up — it is not overheating and throttling, it is simply not running. The temperature must never be read on its own. Always read the clock speed number next to it.

5. "A frame rate drop means a cooling problem, so replace the thermal paste and clean the dust."

Wrong. A frame rate drop triggered by waking is not a cooling problem first. The cooling road described in the guide numbered 136 is about a game that suddenly stutters in the middle of play, and its trigger is not waking. Until the clock speed comparison has been run, the card should not be opened for cleaning. Dust and paste are real causes of real faults, but they are the wrong first suspect for a fault that appears only after waking.

6. "Installing the newest driver will definitely fix it."

Wrong. A forum report on Tom's Hardware describes a case where the problem was precisely a driver version from a particular few months: waking left the core clock soft-locked at 150 MHz, and installing an older driver version was what restored normal behaviour. A driver version is not better simply because it is newer. When a driver version is compared, the version number has to be written down, only one variable may be changed at a time, and updating the driver must not be presented as a universal fix. A Microsoft Q&A reply suggests updating the driver, and that advice is right when the cause is a damaged or outdated driver. The forum case shows the opposite cause: a driver version with a wake defect of its own. Both can be true, because the cause is different in each case.

Comparison Table: What Each Result Points At

ObservationMost likely layerNext test
Frame rate returns at once after restarting only the gameGame and application stateDisplay mode comparison, step 4
Frame rate stays halved after restarting the gameDriver or power stateClock speed comparison, step 2, then driver reset, step 5
Core clock locked low, temperature normalDriver power stateDriver reset comparison, then restart comparison
Core clock normal, frame rate still halvedGame layer or another layerRange comparison across games, then display mode comparison
Only one game slow after wakingThat game's own stateGame restart comparison in that game, then display mode comparison
All games slow after wakingDriver or power stateRestart comparison, step 6
Restart restores speed, shut-down-and-switch-on does notWindows power and Fast StartupFast Startup comparison, keeping one variable changed at a time
Discrete card usage at zero on a laptopDiscrete and integrated graphics switchingCheck which card the game is using, and whether the discrete card is ever called on
Restart does not restore speed eitherNot a wake layer faultFollow the guide numbered 136, the five-path check

What the Sources Disagree About

The sources for this guide do not all say the same thing, and the disagreements have to be written out rather than settled silently.

Restarting the game: helpful or not? A guide on w3tutorials says that in the common case, restarting the game is almost no help and restarting Windows is what fully restores performance. This guide says that restarting the game is a decisive test. The two are not in conflict. The common case is a driver or power state fault, which restarting the game cannot touch. The minority case is a game layer fault, where restarting the game restores speed at once. This guide states the distinction plainly: restarting the game must not be written as a universal fix, and it must not be written as always useless either.

New driver or old driver? A forum report on Tom's Hardware says the problem was precisely a new driver version, and installing an older version was what restored normal behaviour. A Microsoft Q&A reply suggests updating the driver. Both are right, because the cause is different in each case: updating helps when the driver is damaged or outdated, and rolling back helps when the driver version itself has a wake defect. A comparison has to record the version number and change only one variable at a time.

Shut down and switch on, or restart? A guide on w3tutorials states plainly that the two are not the same thing, because Fast Startup only affects shutting down, not restarting. When a reader says "I already restarted and it did not help", the first question has to be whether they really restarted, or whether they shut the computer down and switched it on again. Otherwise the comparison result itself is unreliable.

When to Stop Testing

Stop testing and bring the records for inspection if any of the following apply. The frame rate drop comes with a blue screen, a black screen, a driver crash message, or a burning smell. All seven steps have been run and the layers still cannot be separated. The frame rate stays halved after a real restart, which means the fault is not in the wake layer and belongs with the five-path check in the guide numbered 136. The driver disappears after waking and has to be reinstalled before it returns — that is the shape covered in the guide numbered 494, where the symptom is a missing driver, not a halved frame rate, and it needs a different road. Waking brings a black screen with no picture at all — that is a signal layer problem, and it follows the road in the guide numbered 151. In every case, the written record is what makes the inspection useful: the frame rate before and after waking, the clock speed numbers, the game restart result, the driver reset result, and the restart result, all with the game name, the display mode, and the power plan attached.

What to Bring for Inspection

Bring the computer, or at least bring the written record, and bring the numbers rather than the conclusions. The frame rate before sleep and after waking, in the same game and the same scene. The core clock speed, power draw, and temperature readings from GPU-Z or HWiNFO. The result of the game restart comparison, the driver reset hotkey comparison, and the restart comparison, each with the exact step that was run. The game name, the display mode, and the power plan. The driver version number, and whether Fast Startup is switched on. The published three-tier inspection and repair pricing for this service is HK$400–2,800, and the exact tier depends on what the inspection finds; no price is quoted before the inspection, because the fault has not been located yet. The shop is at Room 1209, Dali Plaza, Shek Mun. The telephone number is 2186 8010, and the WhatsApp number is 6558 6806. There is no shop in Mong Kok, and there is no on-site service.

Frequently Asked Questions

1. My game runs at half speed after the computer wakes from sleep. Is the graphics card failing?

Not necessarily, and usually not. Slowness that appears only after waking points at a state that was not restored properly — either the game did not recover its display mode, or the driver left the card at a low-performance state. Run the game restart comparison first: close the game and reopen the same game in the same scene. If the frame rate returns at once, the fault is in the game layer. If it stays halved, the fault is below the game, and the clock speed comparison is the next step. Do not condemn the hardware before both comparisons have been run.

2. Restarting the game fixed it. Does that mean the game is repaired?

No. Restarting the game is a test, not a repair. It proves that the fault belongs in the game layer, because a clean start gave the game a new display mode and a new device context. It does not prove that the fault will never return, and it does not identify which part of the game's state failed to recover. The next step is the display mode comparison: switch between exclusive fullscreen and borderless window, re-select the display and the resolution, and see whether the game remembers its settings on the next start.

3. Restarting the game did nothing. What does that tell me?

It tells you a great deal. A game restart that changes nothing is evidence, not a failure. It proves the fault is below the game, in the driver or power state layer, because restarting the game cannot reach that layer. The next step is the clock speed comparison: open GPU-Z or HWiNFO and look at the core clock while the game runs. A clock locked at a low level with a normal temperature confirms the driver power state layer, and the road from there follows the guide numbered 284.

4. My temperature is normal. Does that mean the card is fine?

No. A normal temperature only shows that the card is not hot. It says nothing about what clock speed the card is running at. A forum user report describes a core clock locked at 658 MHz with the temperature at only 50 degrees Celsius: the card was cool precisely because it was doing so little work. Always read the clock speed number next to the temperature. A card that has never been woken up properly runs cool, and cool is not the same as healthy.

5. Is shutting down and switching on again the same as restarting?

No. With Fast Startup switched on, shutting down is a partial hibernation: the kernel memory image, including the loaded kernel-mode drivers, is written to the hibernation file and loaded back at the next switch-on. A restart performs a complete boot cycle, and Fast Startup does not apply to a restart. If a restart restores the frame rate while shutting down and switching on again does not, the fault points at the Windows power and Fast Startup layer. Always use a real restart for the restart comparison.

6. Will pressing Win + Ctrl + Shift + B fix the problem?

No. That combination restarts the display driver, and the screen goes black for a moment and then returns. It is a comparison, not a repair. If the frame rate returns at once after the hotkey, the result points at the driver state layer. If nothing changes, move on to the restart comparison. A driver state that has to be cleared by hand after every wake is a driver state that has not been restored properly, and the question of why it fails to restore is still open.

7. Only one game is slow after waking. Is that a different problem?

It is the same diagnosis, with the range comparison pointing at the game itself. One game slow leans toward that game's own state, and the game restart comparison will usually confirm it. All games slow leans toward the driver or power state layer, because the fault sits below every game. Run the game restart comparison in the affected game first, and then run the same test in a second game to confirm the range.

8. My laptop game runs at half speed after waking. Could it be running on the wrong graphics processor?

Yes, and it is worth checking before anything else is changed. On a laptop with two graphics processors, the game may run on the integrated graphics after waking because the discrete card was never called on. Look at which card the game is using in Task Manager or in the game's own display settings, and check whether the discrete card's usage is sitting at zero while the game runs. If the discrete card is at zero, the fault is in the switching, not in the card.

9. Should I install the newest driver?

Not automatically. A newer driver is not better simply because it is newer. A forum report describes a case where the problem was precisely a driver version from a particular few months, and installing an older version was what restored normal behaviour. A Microsoft Q&A reply suggests updating the driver, and that advice is right when the cause is a damaged or outdated driver. Record the version number, change only one variable at a time, and do not present updating the driver as a universal fix.

10. The frame rate is still halved after a real restart. What now?

Then the fault is not in the wake layer, and this guide's seven steps have done their job by proving that. The investigation should follow the road in the guide numbered 136, which works through five paths — cooling, power supply, driver, settings, and hardware — for a game that stutters or drops frames suddenly in the middle of play. Bring the written record with you: the frame rate before and after waking, the clock speed numbers, and the result of every comparison. The published three-tier inspection and repair pricing is HK$400–2,800, depending on what the inspection finds. The shop is at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806.

Sources

Related articles

Related services

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