GPU Driver Gone After Waking from Sleep? Sleep-Wake, Link Power Management & Driver Diagnosis Guide

Key Summary
The machine wakes from sleep and the graphics driver is gone — the card is missing, has become the Microsoft Basic Display Adapter, or carries a yellow warning. This guide separates the wake path from the driver layer, using a restart comparison, a sleep-mode comparison, and the Device Manager state to tell the link, power, and driver layers apart.
The Shape of the Problem: Asleep, Then the Driver Is Gone
This guide covers one specific shape, and the shape itself is the whole diagnosis. The machine goes to sleep. It is woken up — by the keyboard, the mouse, the lid, or the power button — and the display comes back, but something is wrong with the graphics card. Device Manager no longer lists the card at all, or the card has become the Microsoft Basic Display Adapter, or it sits there with a yellow warning sign and an error code. The screen resolution may have dropped, and the graphics control panel may be missing. Restarting the machine, or reinstalling the driver, brings the card back — until the next sleep.
That combination is what makes this symptom different from its neighbours. A driver that disappears every time the machine is switched on from cold is a different shape, and it belongs with the boot-time driver diagnosis. A screen that stays black after waking, with the display signal missing, is a different shape too, and it belongs with the wake-black-screen diagnosis. This guide is about the third combination: the trigger is waking from sleep, and the object that disappears is the driver or the device itself in Device Manager.
The first conclusion most owners reach is that the card has failed, and the first remedy most reach for is reinstalling the driver. Both of those first reactions overlook the single fact that defines the symptom: the card works again after a restart. A card that has genuinely failed does not usually come back to life with a restart and then disappear again only after the next sleep. And a driver reinstall that has to be repeated after every sleep is not a repair — it is a workaround that keeps being repeated. The question is not whether the card is dead. The question is which part of the wake path — the sleep mode, the link power management, the device power setting, the fast-startup state, or the driver itself — fails to bring the card back when the machine wakes.
One honest gap must be stated up front. No single authoritative guide isolates this exact shape — a driver disappearing after waking from sleep — as its own diagnostic flow. The method in this guide is assembled from standard practices: the device-state table used for cards that are not detected, the restart comparison, the power-management settings, and the sleep-state reporting tools. It is presented as a diagnostic assembly, not as official guidance, and the sleep-wake testing in it should not be read as an official procedure.
Why Waking Up Is Not the Same as Starting Up
The key to this symptom is that waking from sleep and starting from cold are not the same operation, even though both end with a working desktop. Starting from cold runs the full power-on sequence: the hardware is enumerated from scratch, the driver is loaded fresh, and every device is brought up in order. Waking from sleep has to restore a state that was frozen in place: the link between the card and the system has to be retrained, the device has to be powered back up in time, and the driver has to be reloaded or resumed — all within the timing the wake sequence allows.
That separation is why a card can be perfect at cold start and missing after a wake. Something can go wrong in the restore step while the cold-start step keeps working, because the restore step is the one that exercises the link retraining, the power-state transition, and the driver's resume path. A fault in any of those three can leave the card unlisted, unloaded, or failed — while a cold start, which never takes that path, keeps looking fine.
There is one more point to settle before the layers begin. Restarting the machine and watching the driver come back proves something precise, and it proves nothing beyond that. It shows that the card and the driver can still work, and that the fault lives in the wake path. It does not prove the driver is broken, it does not prove the card is broken, and it does not repair anything — the next sleep can produce the same disappearance again. The restart is a comparison test, and the rest of the guide is about narrowing the wake path down, one variable at a time.
The Five Suspicion Layers
A driver that disappears after waking can fail at five layers, and they must be kept separate. Changing the power setting, the driver, and the sleep mode all at once and finding the card back proves nothing about which change did the work, and the driver will disappear again the next time only one of those conditions is present. The layers below run from the most direct test to the hardest to isolate, and each one has its own deciding evidence.
| Layer | What it covers | Deciding evidence |
|---|---|---|
| 1. Sleep-mode layer | Traditional sleep (S3) against Modern Standby (S0 low-power idle) — two different power models | Only one sleep mode produces the disappearance |
| 2. Link power management layer | PCIe link-state power management (L0s and L1) and Active State Power Management | Turning link-state power management off changes the result |
| 3. Device power management layer | The device power setting and the driver power-management mode that wake the card | The card returns when the device is not allowed to power down |
| 4. Fast startup and hibernation-state layer | Fast startup as a hybrid shutdown, and a damaged state carried in the hibernation file | A restart clears it while a shutdown does not |
| 5. Driver and Windows Update layer | Driver compatibility with sleep-wake, a driver replaced behind the scenes by Windows Update, and damaged driver files | A clean driver reinstall changes the symptom |
Layer One: The Sleep-Mode Layer
The first layer is that "sleep" is not one thing. Traditional sleep, known as S3, suspends the machine to memory and powers most of the hardware down. Modern Standby, known as S0 low-power idle, keeps the system technically running in a low-power idle state, with the driver still loaded. The two are different power models, not two names for the same sleep, and they are mutually exclusive on a given machine — a platform that uses Modern Standby generally does not offer S1 to S3.
The evidence for this layer is which sleep mode the machine actually uses and which one produces the fault. The reporting tool for this is `powercfg /a`, which lists the sleep states the machine supports and the one in use — read the report as it appears on your own machine, because the result varies by platform; it is a reading from your own machine, not a general standard. If sleep wakes produce the disappearance and hibernation wakes do not, the fault points at the sleep power-state layer. If the reverse is true, it points elsewhere. Do not try to force a machine from Modern Standby back to S3 by editing the registry or by any other forced conversion: third-party write-ups state plainly that the conversion is a platform choice, not an ordinary Windows setting or a simple firmware change, and forcing it carries risk.
Layer Two: The Link Power Management Layer
The second layer is the link between the graphics card and the system. That link runs over PCIe, and when the machine sleeps the link drops to a low-power state — L0s or L1 under Active State Power Management. Waking up means retraining the link: the two ends have to re-establish communication, renegotiate speed, and bring the device back into the system's device list.
The evidence for this layer is the link power setting. In the power options, the setting reported as PCI Express, then Link State Power Management, can be changed from maximum power saving to off as a comparison, one change at a time. If the card comes back reliably after that change, the fault points at the link power management layer. That change is a comparison test, not a repair: leaving link power management off permanently gives up the power saving it exists to provide, and it is not a fix.
Layer Three: The Device Power Management Layer
The third layer is the power management of the device itself. The graphics card has a device setting — reported in its device properties as allowing the computer to turn the device off to save power — and the graphics driver has its own power-management mode. When the machine sleeps, the card is powered down; when it wakes, the card has to be powered back up and the driver has to resume in time. If the device is not woken promptly, the system can finish the wake sequence without it, and the card is left unlisted or unloaded.
The evidence for this layer is the device setting and the timing of the wake. Check the card's device properties for its power-management setting, and note whether the disappearance follows a particular wake method — keyboard, mouse, lid, or power button — since different wake methods take different paths through the wake sequence. Change only one setting at a time, and write down the result after each change.
Layer Four: The Fast Startup and Hibernation-State Layer
The fourth layer is the state left behind by the previous shutdown. Fast startup is a hybrid shutdown: instead of closing everything down, the system writes part of its state into the hibernation file and restores it at the next start. A damaged state can therefore be written into that file, restored at the next start, and carried forward — so a fault that looks like a wake problem can actually be a leftover from the last shutdown.
The evidence for this layer is the difference between a restart and a shutdown followed by a start. A restart reloads the system cleanly and does not use the fast-startup path; a shutdown followed by a start does. If a restart clears the fault while a shutdown and start does not, the fault points at the fast-startup state layer. This is also why the restart comparison in the deciding fork below must be a genuine restart, not a shutdown followed by a start — the two are not the same operation, and confusing them destroys the evidence.
Layer Five: The Driver and Windows Update Layer
The last layer is the driver itself. A driver version can sit badly with the sleep-wake path even while it works perfectly at cold start, because the resume path exercises different driver code from the load path. Windows Update can also replace the driver behind the scenes, so the version installed today may not be the version that was installed yesterday — a driver that disappears after a wake can therefore be a driver that was silently changed overnight. Damaged driver files sit in this layer too.
The evidence for this layer is a clean driver reinstall, carried out as a comparison test and not as a repair. A clean-removal tool such as Display Driver Uninstaller, used in Safe Mode before reinstalling, is the usual way to make that comparison clean — and the replacement driver should be downloaded first, with the network disconnected during the reinstall, so that Windows Update cannot quietly install an older driver in the middle of the test. A clean reinstall that changes the symptom points at the driver layer; one that changes nothing clears it. Do not write any driver version number down as a general standard — versions change, and the current official release is the reference. And do not write a switch between driver branches as a fix: it is a comparison test, and a branch change that clears the symptom proves the branch matters for this machine, not that one branch is correct for every machine.
The Deciding Fork: Restart Immediately, Then Read Device Manager
The whole guide turns on two tests, and they should be done before any setting is changed.
First, the restart comparison. The moment the driver is missing after a wake, do a genuine restart — using the Restart option, not shutting down and starting again. If the driver comes straight back after the restart, the fault is locked to the wake path: the link or the power state failed to restore. If the restart does not bring it back either, leave the wake layer behind and move to the driver and Windows Update layer.
Second, the Device Manager reading, done before the restart. Open Device Manager and note exactly what is there, because the three possible states point in three different directions and must not be mixed up. If the card is completely missing — not even listed as a hidden device — the fault points at the enumeration or link layer. If the card has become the Microsoft Basic Display Adapter, the hardware is recognised but the manufacturer's driver has not loaded, which points at the driver-not-loaded layer. If the card is listed with a yellow warning sign and an error code, such as Code 43, 31, or 45, the driver tried to load and failed, which points at the driver-load-failure layer. The three readings are different evidence, and they lead to different investigations.
Inside the wake path there is a second fork, on the kind of sleep. Try sleep, hibernation, and a shutdown followed by a start, on the same machine, and note which one produces the disappearance. If only sleep wakes produce it, the fault points at the sleep power-state layer. If only hibernation wakes produce it, it points at the hibernation-file and fast-startup layer. If a shutdown followed by a start produces it too, this is not a sleep-wake shape at all, and the diagnosis belongs with the boot-time driver guide.
A third fork sits in Device Manager itself. Choose the option to show hidden devices, and look for the card as a greyed-out entry — a greyed-out entry is the record of a device that is not currently active. Then use the option to scan for hardware changes. If the card returns after the scan, the fault points at the enumeration layer. If nothing returns, the investigation continues to the power-management and driver layers.
One caution applies to every fork in this guide. The sleep-state and power-reporting tools describe the machine they are run on, and their output varies between machines — it must be read as it appears on the reader's own machine, never written down as a general standard. And the sleep-wake testing described here is an assembly of standard practices, not official guidance: no single authoritative guide isolates this shape as its own diagnostic flow, and that gap is stated honestly rather than smoothed over.
What the Published Sources Say, and What They Do Not Prove
The published sources for this guide describe general card-detection problems and sleep models, and none of them can be turned into a guide for this exact shape on its own. They are worth naming honestly, because a reader who only sees one of them can easily mistake a general guide for a sleep-wake guide.
A third-party guide on cards not being detected, re-checked in August 2026, gives a device-state table: the Basic Display Adapter means the manufacturer's driver is missing; a yellow warning means a driver, resource, firmware, or power problem; a card visible in the firmware but not in Windows points at the Windows, driver, or chipset layer; a card visible in neither points at the power, slot, or hardware layer. The same guide advises a complete restart, and specifically shutting down and starting the machine rather than relying only on sleep and wake — direct support for the restart comparison in this guide. It also describes the keyboard shortcut to reset the graphics driver, Win plus Ctrl plus Shift plus B, and explains error codes including Code 22, 28, 31, 43, 45, and 29. That guide is about cards not being detected in general, not about sleep-wake specifically, so only its state table and its restart comparison apply here — its boot-time investigation does not apply to a wake investigation.
A second third-party guide on a card not being detected describes a laptop variant: under one manufacturer's Eco Mode, the dedicated card does not appear in Device Manager at all, and it returns when the mode is changed to Standard. It also notes that the Basic Display Adapter is, in a sense, good news — Windows is running a generic fallback driver, which means the hardware itself is recognised — and it advises downloading the new driver first and disconnecting from the network before a clean reinstall, so that Windows Update cannot quietly install an older driver during the process. The laptop Eco Mode variant is a separate shape from the desktop sleep-wake shape, and the two must not be mixed.
A third third-party guide on a missing card explains that a greyed-out entry in Device Manager is the record of a device that is not currently active, and that showing hidden devices is the documented way to see such records. It gives the meanings of Code 43, 31, and 45, and it notes that a laptop's dedicated card is controlled by the manufacturer's own software package while a desktop card depends on the PCIe slot and its power supply. It is a third-party compilation, not official documentation, and its claims should be read as such.
A fourth third-party write-up, about a laptop found flat by morning, explains Modern Standby as S0 low-power idle — the system technically still running, with the driver still loaded — as opposed to the S1 to S3 states, which are mutually exclusive with it. It describes the reporting tools `powercfg /a` for the sleep mode in use, `powercfg /sleepstudy` for which components were active during sleep, and `powercfg /lastwake` for the last wake source, and it states that the setting cannot be relied on to switch a machine from Modern Standby back to S3. That write-up cites Microsoft documentation for its meaning, but the Microsoft Modern Standby documentation itself was not read directly for this guide — it was reached only through this third-party explanation — so nothing in this guide should be written as a direct quotation from Microsoft. The reader should check the current Microsoft documentation directly.
A fifth third-party guide on checking sleep states with `powercfg /a` describes the same reporting output — Standby (S0 Low Power Idle) against Standby (S3) — and states that a Modern Standby platform generally does not support S1 to S3, because that is a platform choice rather than a Windows fault, and that changing the power model is not something an ordinary Windows setting or a simple firmware change can do. It is likewise a third-party compilation, and it should be checked against the current official documentation.
Three conflicts and gaps in the sources are set out plainly below, because glossing over them would mislead the reader. First, the honest gap: no single authoritative guide isolates a driver disappearing after waking from sleep as its own diagnostic flow. The five sources above cover general card detection or sleep models, and the method in this guide is assembled from standard practices. This guide says so plainly: the sleep-wake testing in it is not official guidance. Second, boot against wake: the boot-time driver guide's five cause chains are all about switching on, and this guide's cause chains are all about waking up. The three-way device-state reading is similar in both, but the trigger is different, and the boot-time conclusions cannot simply be carried across to a wake. Third, black screen against disappearance: the wake black-screen guide describes a watchdog timeout and a display link retraining failure whose visible result is a black screen — a signal-layer fault. This guide describes a device or driver disappearing from Device Manager — an enumeration and driver-layer fault. The trigger is the same, the object is different, and the black-screen explanation does not explain a disappearing driver.
The Seven Steps
Work through these steps in order, changing one thing at a time and writing down the result after each step. A test that changes two things at once proves nothing about which change mattered, and the driver will disappear again the next time only one of those conditions is present.
Step 1: Record the Baseline
Before changing anything, write down which kind of sleep was used — sleep, hibernation, or shutdown — which wake method was used — keyboard, mouse, lid, or power button — what Device Manager shows after the wake — the card completely missing, the Basic Display Adapter, or a yellow warning with its error code — whether the resolution has dropped, and whether the graphics control panel is missing. That note is the baseline every later step is compared against, and it is also the record an in-shop inspection will need.
Step 2: The Restart Comparison
Do a genuine restart immediately — using the Restart option, not shutting down and starting again. If the driver comes straight back, the fault is locked to the wake path. If the restart does not bring it back either, move to the driver layer. Write down the result, and do not skip this step: it is the single test that separates a wake fault from a driver fault.
Step 3: The Sleep-Mode Comparison
On the same machine, try each kind of rest in turn: a sleep followed by a wake, a hibernation followed by a wake, and a shutdown followed by a start. Note which one produces the disappearance. If only sleep wakes produce it, the fault points at the sleep power-state layer. If only hibernation wakes produce it, it points at the hibernation-file and fast-startup layer. If all three produce it, this is not a sleep-wake shape, and the diagnosis belongs with the boot-time driver guide.
Step 4: The Graphics Driver Reset Comparison
When the fault has appeared after a wake, press Win plus Ctrl plus Shift plus B to reset the graphics driver. The screen should flash once. If the picture comes back immediately, the fault points at the driver-loading layer. If nothing happens, keep going down the steps. This is a comparison test, not a repair.
Step 5: The Device Manager Comparison
Open Device Manager, choose the option to show hidden devices, and look for the card as a greyed-out entry — the record of a device that is not currently active. Then use the option to scan for hardware changes. If the card returns after the scan, the fault points at the enumeration layer. If it does not, keep going down the steps.
Step 6: The Power-Management Comparison
In the power options, change the PCI Express link-state power management setting from maximum power saving to off, and try one sleep-wake cycle. Then check the card's device properties for its own power-management setting. Change only one setting at a time, and write down the result after each change. These are comparison tests, not repairs: leaving link power management off permanently gives up the power saving it exists to provide.
Step 7: The Driver Comparison, Then Write the Three-Sentence Conclusion
Do one clean driver reinstall with a clean-removal tool in Safe Mode — a comparison test, not a repair — having first downloaded the replacement driver and disconnected from the network so Windows Update cannot install an older driver during the process. Then watch the machine for several days, noting which kind of sleep and which wake method produce the disappearance. Then write the conclusion in three sentences: what the symptom is, which change made the symptom appear or disappear, and which possibilities are still not ruled out. Bring that note, together with the `powercfg /a` report, the Device Manager state, and the steps already tried with their results, to any in-shop inspection.
Six Common Misconceptions
Misconception 1: "The driver is gone after waking, so the card is dead"
Wrong. A wake is a restore operation, and the link or the power state failing to restore is the most common trigger for this shape. A restart bringing the card straight back is further evidence against a dead card — a genuinely failed card does not usually return with a restart and then disappear only after the next sleep.
Misconception 2: "A restart fixed it"
Wrong. A restart only reloads the driver; it does not repair whatever failed during the wake. The next sleep can produce the same disappearance again. Follow the result of the restart comparison through the sleep-mode and power-management steps instead of treating the restart as a fix.
Misconception 3: "Sleep is just sleep"
Wrong. Sleep in the traditional S3 sense, hibernation in the S4 sense, and a fast-startup shutdown are three different paths, and which one produces the disappearance points at completely different layers. The kind of sleep has to be recorded before any layer is blamed.
Misconception 4: "A black screen after waking and a missing driver are the same thing"
Wrong. The wake black-screen guide describes a signal-layer fault: the display link fails to retrain and the screen stays black. This guide describes a device or driver disappearing from Device Manager: an enumeration and driver-layer fault. A black screen can be a signal problem; a missing driver is never one.
Misconception 5: "It must be a driver problem"
Wrong. Link power management, the enumeration layer, and a damaged fast-startup state can all leave the driver missing after a wake. Until the restart comparison and the sleep-mode comparison have been done, do not settle on the driver as the cause.
Misconception 6: "Turning sleep off solves the problem"
Wrong. Turning sleep off only avoids the path where the fault appears; it does not repair it. Running a machine that never sleeps gives up the power saving and the fast wake that sleep exists to provide, and it is not a fix.
Boundaries: Problems That Look Similar but Are Not This One
Several problems look like a driver disappearing after waking but belong to other guides, and naming them here saves the reader from following the wrong diagnosis.
A driver that disappears every time the machine is switched on from cold belongs with the boot-time driver diagnosis, whose cause chains are all about switching on — Windows Update installing an older driver, fast startup, driver signing, a specific AMD error, and power or slot looseness. A screen that stays black after waking, with the display signal missing, belongs with the wake black-screen diagnosis, which is about the signal layer and the display link retraining. A card that has become the Microsoft Basic Display Adapter without any sleep involved belongs with the basic-display-adapter diagnosis, which is about missing drivers and Windows overwriting them at install or boot time. A machine that needs its DisplayPort cable unplugged and plugged back in at every cold start belongs with the cold-boot handshake diagnosis, where the trigger is switching on and the test is hot-plugging. And a cold first start with no picture that only works after a second restart belongs with the cold-boot diagnosis, which is about power sequencing and memory training, not about a driver disappearing. This guide only covers the shape where the trigger is waking from sleep and the object that disappears is the driver or the device in Device Manager.
One laptop variant deserves its own note. Under one manufacturer's Eco Mode, the dedicated card does not appear in Device Manager at all, and it returns when the mode is changed to Standard. That is a laptop power-mode shape, not a desktop sleep-wake shape, and the two must not be mixed: check the laptop's own power mode before treating the disappearance as a wake fault.
When to Stop and Bring the Machine In
Stop when all seven steps have been worked through, one variable at a time, and the layers still cannot be separated. Stop sooner if any step involves opening the machine or changing firmware, or if the driver problem appears together with artefacts across the whole display, a burning smell, or visible damage — those belong to professional inspection.
Bring the machine in with the notes from the seven steps: which kind of sleep and which wake method produce the disappearance; what Device Manager shows after the wake; the result of the restart comparison; the result of the sleep-mode comparison; the result of the graphics-driver reset; the result of the hidden-devices scan; which power-management settings were changed and what each change did; and which driver versions were compared and whether a clean reinstall was done. That record is what makes an in-shop inspection efficient, because it shows exactly which layers were tested and which evidence each one produced.
In-store inspection takes place at Room 1209, Dali Plaza, Shek Mun. You can call 2186 8010, or reach the shop on WhatsApp at 6558 6806. Inspection is charged at the site's published three-tier fees of HK$400–2,800 (Tier 1 HK$400–780, Tier 2 HK$480–1,380, Tier 3 HK$680–2,800), with the actual price subject to a post-inspection quote. There is no home-visit service and no Mong Kok shop; inspection is carried out in the shop.
A diagnosis that ends at the stop line is not a failed diagnosis. A note that says "sleep wake, card missing from Device Manager, restart brings it back, hibernation wake fine, link-state power management off made no difference, clean reinstall done, still disappears after sleep" is a precise description of where the fault lives, and it is exactly what an in-shop inspection needs to start from the right place instead of from the beginning.
FAQ
I woke my computer and the graphics card is gone from Device Manager — is the card dead?
Not necessarily, and a restart bringing it back is itself evidence against that conclusion. Waking is a restore operation, and the link or the power state failing to restore is the most common trigger for this shape. Do the restart comparison first: if a genuine restart brings the driver straight back, the fault is locked to the wake path, and the card itself is not proven broken.
A restart brought the driver back — does that mean it is fixed?
No. A restart only reloads the driver; it does not repair whatever failed during the wake. The next sleep can produce the same disappearance again. Follow the result through the sleep-mode comparison and the power-management comparison, and watch the machine for several days, noting which kind of sleep and which wake method produce the fault.
Device Manager shows the Microsoft Basic Display Adapter instead of my card — what does that mean?
It means Windows is running a generic fallback driver, which is in a sense good news: the hardware itself is recognised, but the manufacturer's driver has not loaded. That is a different state from the card being completely missing, and it points at the driver-not-loaded layer rather than at the enumeration or link layer. Note the exact state before doing anything else.
My card has a yellow warning sign with Code 43 — what does that mean?
It means the driver tried to load and failed. Code 43, along with codes such as 31 and 45, describes a driver or resource problem, and it is a different state from the card being completely missing or from the Basic Display Adapter. Note the exact code, because the three states point in different directions and must not be mixed up.
The card is completely missing, not even as a hidden device — what should I do?
Open Device Manager, choose the option to show hidden devices, and look for the card as a greyed-out entry — the record of a device that is not currently active. Then use the option to scan for hardware changes. If the card returns after the scan, the fault points at the enumeration layer. If it does not, work through the power-management and driver steps.
Does it matter whether I use sleep or hibernation?
Yes — they are different power paths, and which one produces the disappearance points at different layers. Try sleep, hibernation, and a shutdown followed by a start on the same machine, and note which one produces the fault. If only sleep wakes produce it, the fault points at the sleep power-state layer; if only hibernation wakes produce it, it points at the hibernation-file and fast-startup layer.
How do I find out which sleep mode my machine uses?
Run `powercfg /a` and read the report as it appears on your own machine. It lists the sleep states the machine supports — Standby (S0 Low Power Idle) for Modern Standby, or Standby (S3) for traditional sleep. The result varies by platform — read it as your own machine's report, not a general standard. Do not try to force a machine from Modern Standby back to S3: that is a platform choice, not an ordinary Windows setting.
I pressed Win + Ctrl + Shift + B and nothing happened — does that rule out the driver?
No. The graphics-driver reset is one comparison test, and a lack of response only means the fault is not in the driver-loading layer that the reset exercises. Keep going through the steps: the Device Manager comparison and the power-management comparison test different layers, and the driver comparison is still to come.
Should I turn off link-state power management to fix this?
Only as a comparison test, not as a repair. Changing the PCI Express link-state power management setting from maximum power saving to off and trying one sleep-wake cycle tells you whether the fault points at the link layer. Leaving it off permanently gives up the power saving it exists to provide, so treat the result as diagnostic evidence and note it down.
If all seven steps still cannot separate the layers, what should I bring for inspection?
The notes from the seven steps: which kind of sleep and wake method produce the disappearance; what Device Manager shows after the wake; the restart comparison result; the sleep-mode comparison result; the graphics-driver reset result; the hidden-devices scan result; which power-management settings were changed and what each change did; and which driver versions were compared and whether a clean reinstall was done. The shop is at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806. Inspection is charged at the site's published three-tier fees of HK$400–2,800, quoted after inspection. There is no home-visit service and no Mong Kok shop.