Laptop Wi-Fi Adapter Missing or Intermittently Missing from Device Manager: An Advanced Diagnosis Guide — After the Basic Checks, Separating BIOS, Power, Driver, Slot, and Board Layers

A laptop Wi-Fi card disappearing from Device Manager is one of the easiest faults to misdiagnose. The reader's first sentence is usually 'my Wi-Fi is gone', but behind that sentence can sit four completely different states: completely absent, greyed out and hidden, listed with a yellow warning, or listed but functionally absent. Each state points at a different layer, and each layer has a different fix. This guide assumes the basic checks are already done: a genuine Restart, a flight-mode and physical-switch check, an uninstall and rescan in Device Manager, a driver reinstall from the vendor site, and a Windows Update check. Advanced diagnosis begins only after those basics have failed to resolve the problem.

Article link: https://www.laptops-king.com/en/blog/laptop-wifi-adapter-missing-device-manager-advanced-hk/

1. Scope first: this is about a card disappearing, not about a connection dropping

This guide does not cover 'connected but often disconnecting', weak signal, slow speed, or a single band failing to connect. It covers the case where the Wi-Fi 'node' itself is missing from Windows. Concrete forms include: no Wi-Fi switch at all in Settings, not even the switch itself; a taskbar Wi-Fi icon replaced by a globe icon or 'no connections available'; no wireless card under Network adapters in Device Manager; a wireless card present but semi-transparent, meaning a hidden device, with Code 45; a wireless card present with a yellow exclamation mark, with Code 10, Code 43, or another error code; no wireless card in Device Manager, but an unknown device or Network Controller under Other devices; Wi-Fi present for the first minutes after boot, then gone after sleep, closing the lid, a Restart, or a plugged-in to battery change, returning only after another Restart.

The dividing test is simple. If Device Manager still shows the Wi-Fi card and the trouble is reaching a network, disconnecting, or slow speed, that is a connection problem, not this guide's topic. If Device Manager has no card, or only a hidden item or an unknown device, that is this guide's topic. Stating this division at the start matters, because otherwise the reader will drag router and signal problems into the picture and the misdiagnosis rate will climb sharply.

2. A five-layer model: what has to hold for a card to appear

An internal Wi-Fi card moves from 'absent' to 'present' through a chain, and a break at any link produces the same surface result — 'Device Manager has no Wi-Fi' — while the evidence and the fix are completely different.

Layer	Contents	What happens when it breaks
Layer 0: platform switch	Whether WLAN or Wireless is disabled in BIOS or UEFI; some models have a BIOS allowlist or a slot switch	BIOS never recognises it, so Windows never gets a chance to see it
Layer 1: physical connection and power	M.2 slot contact, screw fixing, antenna leads, board power rails, the PCIe root port in the PCH or SoC	PCIe enumeration fails, so Windows sees no device, or sees it but cannot start it
Layer 2: bus enumeration	BIOS or UEFI and the Windows Plug and Play manager scanning PCIe devices at boot and creating a device node	Enumeration failure leaves not even an unknown device; enumeration success with a resource or firmware problem leaves an unknown device or an error code
Layer 3: driver and firmware	The driver, the driver store, firmware, chipset drivers, and the driver version pushed by Windows Update	The device is seen but cannot start, meaning Code 10 or Code 43, or a driver mismatch makes it come and go
Layer 4: power management and standby	PCIe Link State Power Management, the card's own power-saving setting, Modern Standby, and Fast Startup	Disappearance after sleep, closing the lid, or a plugged-in change, returning after a Restart; one of the most common sources of intermittent cases

The key insight is that one sentence — 'Device Manager has no Wi-Fi' — can come from any of the five layers, so diagnosis cannot rest on one symptom alone. It has to exclude layer by layer.

One logic point carries the whole guide: no driver does not mean no device. If a piece of hardware is enumerated over PCIe, Device Manager will generally show it as an unknown device, a Network Controller, or a name with a yellow exclamation mark, even with no driver installed. If Device Manager has no unknown device and no hidden device either, the problem is more likely in Layers 0 to 2 — BIOS disablement, poor slot contact, a power problem, an enumeration failure, a faulty card, or a board-level fault — than in the driver layer. Conversely, a card that is seen but carries Code 10 or Code 43 points more toward Layers 3 to 4. This distinction is the diagnostic backbone of the article, and it must never be written as 'reinstalling the driver fixes everything'.

3. Intermittent versus permanent: two faults, two lines of thought

Feature	Intermittent disappearance	Permanent disappearance
Definition	Sometimes seen, sometimes not; triggered by Restart, sleep, closing the lid, or a plugged-in change	Long-term absence, whatever Restart or driver reinstall is tried
Common triggers	Sleep or Modern Standby wake, closing the lid, a power-mode change, after a Windows Update, after heat or vibration — as directions for checking only, not asserted causes	BIOS disablement, a loose or damaged card, a damaged slot, a power-rail fault, a PCH-layer fault, a CNVio mismatch that appears only after a card swap
Diagnostic focus	Record the trigger: which action preceded the disappearance? Does a Restart or a full shutdown bring it back? Does it happen on battery or plugged in?	Record the invariants: does BIOS see it? Is there a hidden device? Is there an unknown device? Does a USB adapter work?
Misdiagnosis risk	Mistaken for a router problem, because 'sometimes works, sometimes not' looks like a signal problem	Mistaken belief that reinstalling Windows can fix a hardware-layer problem, wasting time and adding data risk

The reader should first keep a trigger log: before each disappearance, what was done — sleep, closing the lid, an update, moving the machine, plugging in — recorded over three to five events, after which the pattern often emerges by itself. One event is not a pattern, three start to look like one, and ten carry real weight. Records should state facts, not inferences: write 'after closing the lid for ten minutes and reopening it, the card was gone', not 'I think closing the lid broke it'.

Four time patterns for intermittent cases deserve separate recording. Disappearance only after waking points at Modern Standby, power management, or the driver wake path. Disappearance after a Restart, returning after another Restart, points at enumeration timing, a driver-loading race, or unstable firmware initialisation. Disappearance after a period of use points at heat, unstable power, card ageing, or a driver crash. Disappearance after moving the machine or opening and closing the screen points at slot contact, cabling, antenna routing, or board solder joints, as a direction for checking, not an asserted cause.

4. Symptom map: which picture points at which layer

What Device Manager shows	What Settings or the taskbar shows	First reading
Completely absent, hidden view empty, no unknown device under Other devices	The Wi-Fi switch itself is gone, leaving flight mode or Ethernet	This boot's enumeration never recognised the hardware, or the hardware is gated by BIOS or power, or the card or slot layer is at fault; stop spending time on drivers and move to the BIOS check, the USB control test, and the teardown or inspection layer
Greyed out, semi-transparent, Code 45	Much the same as above	Windows recognised this card before and did not detect it on this boot; possible causes are a loose card, a power interruption, a post-wake enumeration failure, or a phantom node left after a driver uninstall
Listed with a yellow warning, Code 10	Wi-Fi may have a switch but cannot connect, or the switch comes and goes	Windows sees the device but driver loading or device initialisation failed; start with a clean driver reinstall, a chipset-driver check, power-management isolation, and a swap test
Listed with a yellow warning, Code 43	Much the same as above	The device itself reported a problem and Windows disabled it to protect the system; if a clean driver install still leaves Code 43, the swap test moves up in priority
Unknown device or Network Controller under Other devices	Wi-Fi switch gone	PCIe enumeration succeeded, but Windows found no suitable driver or driver loading failed; this is good news in one sense, because it proves the card has power and was enumerated, and the problem is more likely in the driver layer
Card listed with no warning, but no Wi-Fi switch in Settings	Wi-Fi switch gone	Possibly a Windows networking service, flight-mode switch, or system-setting layer problem rather than a missing card; check flight mode and a network reset first, and do not jump straight to a card swap

Each error code needs three columns — the Windows meaning, what it does not mean, and the next step — not a one-line definition. Code 10 does not mean the card is certainly dead, nor that the driver version is certainly wrong; it only says 'cannot start'. Code 43 does not mean the driver is certainly wrong either; a hardware report, a driver-to-firmware communication failure, or an error state reported after a sleep wake can all trigger it. Code 45 does not mean the card has physically fallen out; BIOS disablement, power gating, and enumeration failure can all lead Windows to conclude 'not connected'. Code 31 is 'a driver exists but cannot be loaded', while Code 28 is 'no driver found', and their handling directions differ. Working rules for error codes: follow one code at a time, never change several variables together; after each change, record the new state, because a change of code is itself information — for example, moving from Code 28 to Code 10 means the problem has shifted from 'finding a driver' to 'loading or starting'; and the same code can have several causes, while different codes can point at one root cause, since a driver-crash loop can first show Code 43 and later Code 45.

5. Reading Device Manager at an advanced level: more than a yellow mark

First, Show hidden devices: Device Manager, View, Show hidden devices. Normally shown items are currently connected and recognised by Windows; the extra grey items that appear after turning hidden view on are devices Windows installed before and does not detect now, meaning non-present devices. Grey does not mean 'broken'; it means 'present last time, absent now'. A grey Wi-Fi card means the driver package and hardware ID record are still there, and the problem leans toward 'not enumerated this time' rather than 'driver never installed'. If even hidden view shows nothing, this boot's enumeration had no such hardware at all, or the driver record has been cleared. Do not delete grey items wholesale: deleting every grey item can remove devices that are currently unconnected but will be used again, and it does nothing directly to bring the Wi-Fi card back.

Second, Devices by connection versus Devices by type. The default Devices by type view groups devices by kind, which is convenient for finding a card but hides which bus it hangs from. Switching to Devices by connection shows a tree, roughly: the computer root node, Microsoft ACPI-Compliant System, PCI Express Root Complex, a PCI Express Root Port, and the Wi-Fi card's PCIe part, with a separate USB Root Hub branch carrying the Bluetooth part. This view has four diagnostic values: it confirms which path Wi-Fi takes, since the Wi-Fi card should hang under some PCIe Root Port and the Bluetooth part under USB; it exposes upstream trouble, since a warning on the PCIe Root Port itself or on a USB controller means the Wi-Fi disappearance may be only a downstream symptom; it separates virtual adapters, since VPN, Hyper-V, VMware, VirtualBox, TAP, Bluetooth PAN, and WAN Miniport adapters do not hang under a PCIe Root Port; and it carries a CNVio exception, since an Intel CNVio card's Wi-Fi logic partly sits in the CPU or chipset and its presentation differs from a standard PCIe card, so tree position alone must not settle the question. This view is for looking only: never disable an upstream device in it, since disabling a whole PCIe Root Port or USB Root Hub can take a whole group of devices down with it, including keyboard and touchpad.

Third, read the Properties pages one by one, not only the General page. On General, the Device status field is where the error code lives, and the full sentence should be copied, not just the number; if the status says 'This device is working properly' yet Wi-Fi is still missing, the problem is not in driver loading, and the service layer, the switch layer, or the Wi-Fi and Bluetooth split should be examined. On the Driver page, note the provider — Intel, Realtek, MediaTek, Qualcomm, Microsoft, or the laptop brand — plus the driver date and version, and compare the version after every change, since many people believe they 'updated the driver' when Windows actually installed the same version or a generic version from another vendor; a greyed-out Roll Back Driver button means Windows holds no older version, not that downgrading is impossible, only that it must be done manually. On the Details page, Hardware Ids gives the most precise identification, in a form such as PCI, VEN, DEV, and SUBSYS; Device instance path shows which PCIe slot or USB port it hangs from; Power data only shows which power states are supported and is not fault evidence. On the Events page, installation, start, disable, and removal times and results are recorded, and the point is time alignment: which driver install, Windows Update, or sleep wake does the first disappearance line up with? On the Power Management page, if present, look at 'Allow the computer to turn off this device to save power'; many Modern Standby machines have no such page, and that is not a fault. On the Advanced page, 802.11 mode, bandwidth, roaming aggressiveness, and transmit power affect connection quality and band, and they do not make a card disappear from Device Manager, so they should not be changed at random during advanced diagnosis.

Fourth, Event Viewer and the driver-install log. In Event Viewer, in priority order: Windows Logs, System, filtered to the disappearance time; Application and Service Logs, Microsoft, Windows, WLAN-AutoConfig, Operational; Kernel-PnP setup and operational logs; NDIS; and Kernel-Power entries in the System log, used to align whether disappearance always follows a wake rather than to prove a Wi-Fi fault directly. The working method is to record the exact disappearance time first, then read ten minutes either side of it, and not to read red errors only, since yellow warnings and information entries carry value — for example, a driver loading successfully and then being removed shortly afterwards is itself a sequence pointing at a crash-and-recovery loop. If Event Viewer has no related entries at all, that is itself information: the device may never have been enumerated in time for Windows to log it, or the problem may sit in the BIOS or power layer. The driver-install log lives at C, Windows, INF, setupapi.dev.log, and records each driver install in detail; lines starting with an exclamation mark are warnings, and lines starting with three exclamation marks are failures, which should be read first. An oemXX.inf name is only Windows' internal name for a third-party driver package, not a fault in itself, and it must be matched with pnputil output to learn which vendor, device, and version it belongs to. The log is for reading only, never for editing or deleting.

Fifth, other built-in Windows diagnostic commands, all auxiliary rather than decisive. Pnputil lists network-class devices and their states for comparison with Device Manager; pnputil lists installed third-party driver packages to match oemXX.inf names; netsh wlan show drivers shows the current Wi-Fi driver version and supported radio types; netsh wlan show interfaces shows current wireless interface state; PowerShell Get-NetAdapter lists network adapters and states; powercfg lists supported sleep states; powercfg shows the last wake source, which only says which device or timer woke the machine, not that the device caused the Wi-Fi loss; and powercfg sleep-study reporting shows which devices were active during Modern Standby. The shared limit of all of these commands is that they only ask Windows 'what do you see now': if Windows never enumerated the card, the answer will simply be 'nothing', and that cannot prove a hardware fault.

6. The power layer in depth: why waking is the trigger

Power management deserves to come first among the mechanisms, because it is the most common source of intermittent disappearance and also the easiest to mistake for a router problem. Start by separating Restart from Shutdown. Restart performs a full shutdown plus boot, reloading drivers and core state and re-enumerating hardware. Shutdown with Fast Startup enabled is not a full shutdown but a hybrid shutdown: the kernel and driver session are written to a hibernation file and loaded back on the next boot, rather than enumerating hardware from zero. If the driver state was already faulty, hybrid shutdown simply carries the faulty state into the next boot, and the card stays missing. So the first test in diagnosis must use Restart, not shutdown and start. If the card returns after Restart but disappears again after shutdown and start, that pattern itself points at Fast Startup or hibernation state.

The diagnostic use of Fast Startup is to turn it off, then do a shutdown and start, and see whether the card comes back stably. If turning it off removes the problem, do not declare it fixed at once; observe several sleep wakes and shutdown-and-start cycles to confirm the pattern really changed. Turning Fast Startup off makes boot slightly slower, which is a normal cost, not a fault.

Traditional S3 sleep and Modern Standby deserve separate treatment, because this is the section most easily written wrongly and most worth writing deeply. Traditional S3 sleep almost fully stops the system, and waking re-initialises most hardware. Modern Standby, meaning S0 low-power idle, keeps the system in an extremely low-power but partly active state, able to receive mail and update tiles in the background, with the SoC actively managing power, so an individual device's Power Management tab may simply not appear. In Microsoft community answers, Microsoft staff have stated explicitly that on a Modern Standby system, the Wi-Fi card may have no Power Management tab, because power is handled by the SoC rather than by the traditional per-device power setting. The diagnostic consequences are threefold: a missing Power Management tab must not be read at once as a broken driver, since the machine may simply be running Modern Standby; a Modern Standby machine's 'gone after waking' case should first raise suspicion about the SoC power state machine and the driver or firmware wake path, not a single checkbox; and the test must separate 'gone after sleep' from 'gone after a full shutdown and start' — if only waking loses the card while a full shutdown and start always brings it back, the power-management layer is the strongest suspect, whereas if a full shutdown and start also fails to bring it back, the power layer is much less likely.

The card's own power-saving setting, 'Allow the computer to turn off this device to save power', lets Windows cut power to the card when idle or asleep to save power. Normally, waking should restore power and re-enumerate, but if the driver, firmware, or power state machine has a flaw, waking may fail to enumerate, leaving Wi-Fi gone after waking. Clearing that checkbox is a low-risk isolation test: if clearing it clearly reduces the frequency of post-wake disappearance, the power layer is indicated; if it changes nothing, that switch can be excluded. It must not be written as a permanent cure: clearing power saving is a diagnostic step and a temporary mitigation, not proof of the root cause.

PCIe Link State Power Management works on the PCIe link itself, which has different power states, and at maximum power saving the system pushes the link aggressively into low-power states, requiring link re-training on wake. If the card, the board, or the driver handles re-training imperfectly, the link may fail to come back, and the device disappears from Windows until the next full Restart or hardware rescan. Setting it to Off for both 'on battery' and 'plugged in' is another isolation test, especially suited to the 'gone after sleep or closing the lid, back after Restart' pattern. Turning it off costs a little extra power, and it must not be written as something everyone should leave off permanently.

Plugged-in versus battery changes must also be recorded separately: some intermittent cases appear only on battery, or only plugged in, because power plans can carry different PCIe Link State Power Management settings, card power-saving settings, and processor power management for the two states. Diagnosis must keep 'plugged in' and 'on battery' apart, not merged.

7. The driver and firmware layer: the old-driver trap, clean install, rollback, and hiding

Driver corruption and driver-store conflicts are the most common driver-layer mechanism. The Windows driver store can hold several versions of the same driver at once, and common conflict scenes include: a user installs one version manually, Windows Update then pushes another version, and the two load alternately; an uninstall without ticking 'delete the driver software for this device' leaves a damaged old driver behind, and the next scan loads the same damaged version again; after a card swap, the old card's driver residue conflicts with the new card's driver; and driver files themselves are damaged by disk errors, an interrupted update, or a third-party cleaning tool deleting files. Symptoms are a card seen but carrying Code 10 or Code 43, or a card returning briefly after uninstall and Restart and then disappearing again.

The correct cleaning logic must be written as a procedure, not as the four words 'reinstall the driver'. In preparation, while network access still exists: copy down the current driver provider, version, and date, plus the hardware ID; download the Wi-Fi driver and the Bluetooth driver from the laptop brand's official support page, since many brands split Wi-Fi and Bluetooth into two packages and both are needed; store the installers on a USB stick or a non-system drive folder, so they remain reachable even when Wi-Fi is gone; and note down the Wi-Fi password, since a later network reset would clear saved networks. In execution: in Device Manager, right-click the Wi-Fi card and uninstall the device, ticking the option to remove the driver software; if Bluetooth shares the same card, handle the Bluetooth item the same way, and do not clear Wi-Fi only and leave old Bluetooth residue behind; after uninstalling, do a genuine Restart, not shutdown and start; after Restart, check Device Manager first, then run Scan for hardware changes; run the previously downloaded OEM driver installer, installing Wi-Fi first and then Bluetooth, without running Windows Update at the same time; after installing, Restart again and compare the driver version to confirm it really changed to the one just installed. In verification, add a sleep-and-wake test and a shutdown-and-start test: a single boot test is not enough, since an intermittent problem must be judged by its pattern.

The Windows Update old-driver or wrong-driver trap is real, but easily oversimplified into 'Windows Update broke the machine'. The Windows Update driver catalogue updates on a different rhythm from the vendor site: sometimes the pushed version is older than the vendor site's latest version, and sometimes it is a generic version that does not fully match a particular laptop brand's custom settings, such as antenna configuration, power limits, or regional rules. The result is a machine that was fine before the update and starts disappearing intermittently afterwards, or goes straight to Code 10. Diagnostic clues are a start time matching some Windows Update or driver update, a driver version and date visible on the Driver page, and a Roll Back Driver button that can be clicked, meaning an older version is still held. The handling principle is to use Roll Back Driver or a manual install of the vendor's correct version as an isolation test first, rather than disabling all of Windows Update at once. If a particular driver version is confirmed as the problem, pausing automatic updates for that driver can be considered, but the method varies by Windows version and must not be written as one fixed procedure for all versions. Windows Update must not be written as a universal villain: if the problem predates the update, or if swapping in a USB adapter also disappears, then the driver version is not the problem.

Rollback plus hiding is the complete procedure: rolling back alone is not enough, because the next Windows Update check is very likely to install the same version again and the problem will recur. Microsoft's Show or Hide Updates troubleshooter can hide a specified driver update, so Windows Update temporarily stops offering that version. If the vendor later releases a fixed version, the same tool must be run again, choosing to show hidden updates and unhiding it, before the new version will arrive. Hiding is a temporary block on one version, not a permanent block. Blocking all driver updates through Group Policy or device installation settings is broader and needs care: it blocks every driver update delivered through Windows Update, not just one version, and it also blocks later fixed versions and security driver updates, so it fits only as a short-term diagnostic or transitional measure, not a long-term setting.

The three driver sources must be kept apart. Laptop-brand OEM drivers are tested against the model and usually carry brand custom settings, with the highest fit to BIOS, but their versions may be older and slower to update. Windows Update drivers arrive automatically and are usually signature-verified, but they may be generic versions without OEM custom settings, and their versions may not fully match the machine's BIOS. Chip-vendor generic drivers are usually the newest and fastest to fix, but they may lack OEM custom settings, some functions may be missing, and installing the wrong CNVio or standard version leads directly to Code 10. The working advice is: for laptop Wi-Fi problems, try the OEM driver first; if a clean OEM install still fails, try the chip-vendor generic driver only after confirming the model and interface match exactly; and treat Windows Update drivers as a fallback for 'is there a driver at all', not as the main variable during diagnosis, since their change timing cannot be controlled and comparison testing becomes difficult.

Driver and firmware are two different layers: the driver is software running in Windows that talks to the hardware, while firmware is lower-level code in the card itself or in the board BIOS that initialises the hardware. Driver problems usually mean the device is seen but cannot start; firmware or BIOS problems mean the device may never be enumerated at all, or is enumerated but behaves abnormally. BIOS or UEFI firmware comes from the laptop brand, and updating it carries higher risk — a power loss during update can brick the machine — so readers must not be told to update BIOS before other causes are excluded. Chipset drivers matter too: many people update only the Wi-Fi driver and ignore the chipset driver, yet the PCIe root port, PCH power management, and USB (which is the Bluetooth path) all depend on the chipset driver. If the chipset driver is old or damaged, PCIe enumeration and power-state transitions can both go wrong and indirectly make the Wi-Fi card disappear. The suggested order is to confirm the chipset driver is the laptop brand's correct version first, and then handle the Wi-Fi driver; reversing the order makes test results hard to read.

Secure Boot needs a clear statement of what is relevant and what is not. Secure Boot is a UEFI security mechanism that allows only signed boot loaders and drivers to load at boot. If a driver is unsigned or test-signed, Windows may refuse to load it with Secure Boot on, producing Code 31 class load failures. But Secure Boot itself does not make a normal card with a correctly signed driver disappear from Device Manager, and having Secure Boot on is not a cause of Wi-Fi disappearance, so it must not be turned off 'just to try'. Turning Secure Boot off lowers boot security protection and must not be a routine step in Wi-Fi diagnosis. Normally, Secure Boot should stay in its original state and must not be toggled for a Wi-Fi problem.

The correct driver-handling order runs from lightest to heaviest, with a result check after each step: first, enable plus scan, the lightest; second, disable and re-enable, which re-initialises the driver without deleting driver files; third, update the driver, after confirming the current provider, version, and date and then choosing a source; fourth, roll back the driver, only when the problem appeared immediately after some driver update; fifth, uninstall plus Restart, without ticking delete driver, letting Windows reinstall from the already installed driver package; sixth, a clean install, uninstalling with driver deletion and then installing the OEM driver by hand; seventh, network-stack resets, which fit 'card present but connection abnormal' and will not bring back a card that is fundamentally absent; eighth, a network reset, the heaviest, only after all targeted steps have failed, and only after recording Wi-Fi passwords, VPN settings, and fixed IP or DNS settings; ninth, stop and hand it over for inspection, rather than continuing to more aggressive methods such as reinstalling Windows, flashing BIOS, or editing the registry.

8. The BIOS and UEFI layer: which names to look for, and which do not matter

The laptop BIOS or UEFI can independently switch the internal wireless device on or off. Common location names vary by vendor, and roughly include: Advanced, Onboard Devices Configuration, WLAN or Wireless LAN; Configuration, Wireless; Security, I/O Port Access, Wireless Network; and on some business models, separate Wireless WAN or Bluetooth switches, which must be checked separately rather than assuming Bluetooth on means Wi-Fi on. If the BIOS layer disables it, the Windows Plug and Play manager never receives the device at all, and Device Manager naturally shows nothing. In that case, reinstalling drivers, reinstalling Windows, and changing routers are all useless, because the problem sits below Windows.

Why would it be disabled for no apparent reason? A user or technician changed a BIOS setting and forgot; a BIOS update reset settings to defaults, and the defaults on some models do not enable everything; BIOS settings became corrupt and need Load Defaults followed by item-by-item confirmation; or an enterprise or school model has a management policy locking wireless devices, in which case IT should be asked first and the lock must not be bypassed on one's own.

Checking points are: before entering BIOS, note the machine's brand and model, because every vendor's BIOS interface and keys differ, and no single key should be written as a universal tutorial; once inside BIOS, look for wording such as WLAN, Wireless, or WLAN Device, and confirm it is Enabled; if no such option can be found, that does not mean there is no problem, since some thin models have extremely simplified BIOS interfaces that simply do not show a wireless switch, and the exclusion procedure should be followed instead of searching BIOS to death; after any change, Save and Exit, and never just power off; and if a Load Defaults option exists and no other important setting is known to have been changed, such as boot order or Secure Boot, Load Defaults can be considered followed by re-confirming the wireless switch, but this step needs care and must not be pressed blindly.

The BIOS allowlist needs careful wording: some older models accept only vendor-approved card models, and a non-approved card may fail to boot or boot without being recognised. The phenomenon exists, but whether it applies to a given machine can only be settled against the official service manual, so it must neither be written as a universal rule nor as non-existent. Compatibility must be checked before buying a card, not by buying first and trying.

A BIOS information page that does not list the Wi-Fi card cannot by itself prove the card is dead. Many laptop BIOS pages never list a Wi-Fi card model at all, only major hardware such as CPU, memory, and storage. What carries diagnostic value is whether the wireless enable switch in BIOS is Enabled, not whether the information page lists a model. A BIOS update is not itself a cure for Wi-Fi disappearance, unless the official update notes explicitly state a fix for wireless enumeration, power management, or compatibility with a particular card model. An unmotivated BIOS update carries risk: a failed update can leave the machine unbootable, a far worse outcome than Wi-Fi disappearance alone, and after an update all BIOS settings may revert to defaults, including the wireless setting, boot order, and Secure Boot state, which must then be checked one by one.

9. M.2, PCIe, and CNVio: the easiest place to buy the wrong card

Internal laptop Wi-Fi cards almost all use the M.2 2230 size, meaning 22 mm wide and 30 mm long. But 'the size fits' does not mean 'insert it and it works'; the Key, meaning the notch position, and the interface type must also be checked. Common combinations are: M.2 Key E, the slot designed for wireless cards, supporting PCIe plus USB; M.2 Key A/E, a card with two notches that fits a Key E slot, common on Intel AX200 and AX210 PCIe cards; and M.2 Key E cards with only the E notch, such as Intel AX201 and AX211 CNVio2 cards, which fit a Key E slot physically but may not match in protocol. The article must state plainly: physically insertable does not mean electrically and protocol compatible.

Intel's CNVi architecture differs fundamentally from a traditional PCIe wireless card. A traditional PCIe card carries the whole Wi-Fi plus Bluetooth function — MAC, processor, memory, radio — on the card itself, so the card is a complete device. In the CNVi architecture, most of the digital logic — MAC, processor, memory, and related firmware — moves into the Intel CPU or PCH, and the card carries only the radio-frequency and signal-processing part, called a Companion RF (CRF) module. According to Intel's official support article 'What Are the Intel Integrated Connectivity (CNVi) and Companion RF (CRF) Module?': CNVi consists of the CNVi part inside the processor plus a CRF module in M.2 form; a CRF module can only be used in Intel processor, chipset, or board systems specially designed to support CNVi; even though a CRF can physically be inserted into a standard M.2 Key E slot, it is only compatible with systems designed for CNVi; and the document explicitly advises contacting the system or board vendor to confirm compatibility.

The generation table, compiled from the ASRock official FAQ and Intel community material, is as follows:

Generation	Representative models	Platform direction
CNVio	Intel Wireless-AC 9560, 9462	Intel 300-series chipsets, eighth- and ninth-generation processor era
CNVio2	Intel Wi-Fi 6 AX201, AX211, AX411, AX101, AX203	Intel 400 series and later, depending on model
CNVio3	Intel BE201, BE401, Wi-Fi 7 era, Core Ultra platforms	Only specific new platforms, not backward compatible

The key incompatibility rules are: a CNVio card cannot be used as a CNVio2 card, or the reverse. An AX201 (CNVio2) cannot replace a 9560 (CNVio) in a system that only supports CNVio, even though the physical slot is the same. Model tail digits give a hint but not an absolute rule: Intel AX200 and AX210 ending in '0' are mostly PCIe cards, and AX201 and AX211 ending in '1' are mostly CNVio2 cards. But readers must be told that tail digits are only a first clue, and the final check must be the System Interface Type field in Intel ARK or the vendor specification table, never tail digits alone. PCIe cards such as AX200 and AX210 work in many CNVio2 systems, because a CNVio2 slot usually also supports PCIe signalling; but the reverse — a CNVio2 card in a system that only supports PCIe — mostly fails to be recognised. That asymmetry is the most common source of buying the wrong card.

The compatibility matrix, in conceptual form and always to be rechecked against official specifications before buying, is: standard cards (AX200, AX210, BE200) in a CNVio2-supporting slot usually work, because the slot itself supports standard PCIe signalling, but OEM allowlist or BIOS limits are exceptions, so this is not a complete guarantee; CNVio2 cards (AX201, AX211) in a standard slot do not work, because the card lacks complete logic and depends on the CPU or chipset part that a standard slot does not have; a CNVio2 card in a first-generation CNVio system does not work, because CNVio and CNVio2 are incompatible; and a CNVio3 card (BE201) in a CNVio2 system does not work, since CNVio3 is not backward compatible.

For the disappearance fault, the meaning is: if the machine was fine before, and disappearance started only after a card swap, or the card is directly absent after a swap, the first suspicion should be a CNVio generation mismatch, not a faulty new card. If the machine had no card swap but disappearance started after a BIOS update, the suspicion should be that the BIOS update changed CRF support or enumeration behaviour, and this must be labelled as a not-fully-verified inference, not written as a conclusion.

Socketed versus soldered-down must be separated. The Intel CNVi document notes two CRF forms: 2230 socketed and 1216 soldered-down. The 2230 socketed form is fixed by a screw in an M.2 slot and can be removed, reseated, and swap-tested. The 1216 soldered-down form is soldered directly to the board, cannot be removed or swapped, and is common in ultra-thin machines; if its radio part fails, the repair path is not 'swap the card' but board-level work or a USB Wi-Fi workaround. Readers should be taught to tell them apart by checking the laptop brand's service manual or teardown guide for whether the wireless part is socketed, and must not dismantle the machine before checking the manual; for a soldered-down model, 'replace the card yourself' must not be offered as advice, and the correct wording is that this kind of machine is not suitable for self card replacement and needs assessment by repair staff for board-level or external options.

The PCIe enumeration-failure mechanism is: at boot, BIOS or UEFI scans the PCIe bus, asking each device in turn who it is and how many resources it needs, in a process called enumeration. Enumeration can fail because the device has no power, such as a power-rail fault or a power-management IC problem; because the device has power but does not respond, such as a faulty card or stuck firmware; because of signal-integrity problems, such as poor slot contact, oxidation, looseness, or damaged board traces; because the root port itself is faulty, meaning a PCH-layer fault; or because BIOS enumeration logic is flawed. The result of enumeration failure is that Windows sees no device at all, not even an unknown device. At that point, any driver-layer operation is meaningless, because Windows has no node to bind a driver to. The diagnostic dividing line is: enumeration failure means Device Manager shows nothing at all, including hidden devices and Other devices, while driver failure means Device Manager has a node but with an error code. That dividing line decides whether to keep working in Windows or to move to dismantling or inspection.

On antenna leads, an internal Wi-Fi card usually has two leads: MAIN (primary) and AUX (auxiliary). The leads use tiny coaxial connectors, which are extremely small, fragile, and limited in insertion cycles. The important mechanism clarification is: a loose or broken antenna lead usually causes weak signal, slow speed, or a poor band, and does not directly make the card disappear from Device Manager, because the antenna is a radio path, not a PCIe enumeration path. The exception to state is: if a loose lead shorts, or if dismantling tears off the antenna socket or damages the card or board, the card or enumeration can be damaged indirectly, but that is dismantling damage, not the normal failure mode of the antenna itself. The article must separate the two clearly: an antenna problem is not a disappearance problem, and if the reader's main complaint is disappearance, the antenna must not be listed as the first suspect; but during a teardown check, antenna state should still be recorded together, because antenna connector state affects signal performance after a card swap. Practical notes for antenna connectors are: press or pull vertically and gently, never pull sideways, since sideways pulling easily tears the whole socket off; do not swap the two leads, since MAIN and AUX share the same physical connector shape but swapping them affects MIMO performance, so photograph the routing before dismantling; and route antenna leads away from high voltage, heat, and hinge positions, since long-term flexing at the hinge is a common ageing position for antenna leads, as a direction for checking, not an asserted cause. The colour trap must also be stated: many people assume a black lead must be MAIN and a white lead must be AUX, but lead colours vary by brand, model, and production batch, and must never be judged by colour alone; the correct method is to count connectors and leads and compare with the service manual wiring diagram. The single-antenna trap is: if the machine left the factory with only one antenna lead, swapping in a 2x2 card still only uses one antenna, so speed and stability do not become true 2x2, and Bluetooth may have to share the AUX antenna, making Bluetooth reception worse.

On size traps, M.2 numbers mean size: 22 mm wide plus 30, 42, or 80 mm long. Laptop Wi-Fi cards are almost all 2230, while 2242 and 2280 are mostly SSD sizes, not Wi-Fi card sizes. Buying the wrong size means it physically cannot be fitted, or the screw position does not line up, and forcing it can press against the board and damage other parts. Buying a Wi-Fi card needs '2230' plus Key, interface generation, antenna connector, and the laptop brand allowlist — all four, with none missing. A Key M slot is an SSD slot, not a Wi-Fi slot, and even if the size looks similar, a Wi-Fi card must not be inserted into an SSD slot, or the reverse. Mini PCIe is an old standard, completely different from M.2, and an old machine's Mini PCIe card cannot be replaced with an M.2 card.

The Wi-Fi and Bluetooth relationship must be explained case by case: most modern internal wireless cards are Wi-Fi plus Bluetooth two-in-one, but their system paths differ, with Wi-Fi usually over PCIe and Bluetooth usually over USB through the M.2 slot's USB pins. So the combinations are: Wi-Fi gone with Bluetooth normal points at the PCIe path or the Wi-Fi part, with the USB path normal, and must not be read at once as the whole card being dead; Wi-Fi normal with Bluetooth gone points at the USB path or the Bluetooth driver; both gone together points more at the whole card losing power, BIOS disablement, a slot problem, or a card fault; and both gone but both returning after a Restart leans toward power management or enumeration failure. This comparison table is a very useful diagnostic tool.

Realtek, MediaTek, and Qualcomm naming is also easy to confuse, and cards must be identified by hardware ID plus the full model on the module, not by a model prefix alone. Realtek 8852BE is Wi-Fi 6, PCIe, 2T2R, and 8852CE is its Wi-Fi 6E three-band version, but not every 8852BE machine can swap directly to 8852CE; antenna, driver, Windows version, and router fit must be checked. The 8821CE is Wi-Fi 5, 1x1, single-antenna, and the 8822CE is Wi-Fi 5, 2x2, dual-antenna. MediaTek MT7921 is Wi-Fi 6 over PCIe, and MT7922 is its Wi-Fi 6E extension, with driver and power-management requirements that are not identical to MT7921, so MT7921 drivers must not be applied directly to MT7922. AMD RZ608 and RZ616 are essentially MediaTek chips under AMD branding, but branding, driver packages, and OEM support page names may use AMD names rather than MediaTek names, so the OEM support page must be consulted. Qualcomm laptop Wi-Fi naming is even more scattered: FastConnect is a platform name, not a card model, and the slot form and driver source for models such as QCNFA765 and WCN685x must be checked machine by machine against the OEM support page, never generalised, and some Qualcomm cards are soldered to the board with no M.2 slot to swap at all.

10. Layered exclusion procedure: from outside in, from software to hardware

This chapter is the operational core of the article. The principle is: after each step, record the result before deciding the next step; never do everything at once, or nobody will know which step was the decisive one.

Step 0: record the baseline, before any change. Laptop brand, model, Windows version (10 or 11), and BIOS version if it can be read; wireless card model, if Device Manager still shows it, or from the laptop brand specification table; driver version number and date; symptoms, permanent or intermittent, which action triggers it, and whether Bluetooth disappears together or only Wi-Fi; and recent changes, meaning Windows Update, driver update, BIOS update, hardware change, dropping, water exposure, or dismantling. This baseline must be compared against at every later step; without a baseline, diagnosis becomes random trying.

Step 1: confirm 'true disappearance' versus 'false disappearance'. Open Device Manager, View, Show hidden devices; expand Network adapters and look for a semi-transparent wireless card; expand Other devices and look for an unknown device or Network Controller; open Settings, Network and Internet, and look for a Wi-Fi item; and check the flight-mode switch, the physical switch, and Fn shortcuts. Result classification: a hidden device leads to Step 3, the Code 45 path; an unknown device leads to Step 4, the driver-identification path; a card with an error code leads to Step 4, the Code 10 or Code 43 path; and nothing at all leads to Step 2, the BIOS path.

Step 2: the BIOS or UEFI layer. Restart, enter BIOS, and look for WLAN, Wireless, or Wireless LAN options, confirming Enabled; if none can be found, do not search to death, but record 'BIOS interface shows no wireless switch' and continue; if anything is changed, Save and Exit, re-enter Windows, and recheck Device Manager; and if BIOS offers Load Defaults and no other important setting is known to have been changed, Load Defaults can be considered followed by re-confirming the wireless switch, but this step needs care. Judgement: BIOS on but Windows still showing nothing excludes BIOS disablement and moves to Step 5, the external control test, and Step 6, the teardown layer; and a BIOS that shows no wireless device information at all is itself a signal pointing at the enumeration layer.

Step 3: hidden-device or Code 45 handling. In Device Manager, right-click the hidden wireless card and uninstall the device, ticking 'delete the driver software for this device' if that option exists; Restart, using Restart rather than shutdown and start, to avoid Fast Startup carrying a stale state; then in Device Manager, Action, Scan for hardware changes; and observe whether the device returns, whether it returns with an error code, or whether even the hidden device is gone. Judgement: returning and normal means recording it and observing for several days, moving to power-management isolation if it disappears again; returning with Code 10 or Code 43 moves to the driver-cleaning path; and not returning at all falls back to the 'nothing at all' path and Step 5.

Step 4: driver and power-management isolation, the last line of software defence, and this step must be split, never changed all at once. 4A, clean driver reinstall: download the correct driver first, laptop brand first and card vendor second, stored on USB; uninstall plus driver deletion, Restart, and manual install of the downloaded driver; then record the version number and observe. 4B, chipset-driver confirmation: confirm from the laptop brand site whether the chipset driver is the latest correct version, and if it is old, update the chipset first and then retest Wi-Fi. 4C, power-management isolation, tested separately: test one, the card Power Management page if present, clearing 'Allow the computer to turn off this device to save power', and observing whether post-sleep disappearance stops; test two, PCIe Link State Power Management set to Off for both plugged-in and on-battery, and observing; test three, temporarily disabling Fast Startup, doing a full shutdown and start, and observing. Between each test, allow at least one sleep-and-wake cycle, or one to two days of normal use, before judging improvement. Changing all three at once means that even if it improves, nobody knows which change mattered. 4D, Windows Update driver version comparison: check the driver version and date against the problem start time; if the start time matches some driver update, try Roll Back Driver or a manual install of the older correct version; and record which version is faulty and which version is normal, since that information matters greatly to repair staff. Judgement: if everything in the software layer has been tried and the problem remains, stop circling in software and move to Step 5; if one software step clearly improves things, record which step, and remind the reader that improvement is not a cure, and that a recurrence means running the full exclusion again.

Step 5: the USB Wi-Fi adapter control test, the key watershed. The principle is simple: internal Wi-Fi runs over PCIe plus board power plus BIOS enumeration, while a USB adapter runs over the USB bus and completely bypasses the internal card's PCIe path, M.2 slot, and BIOS wireless switch. Method: insert a known-good USB Wi-Fi adapter, borrowed if needed with no need to buy one; install its driver, which Windows usually installs automatically or the vendor supplies; and observe whether the USB adapter can scan Wi-Fi, connect, and browse normally. Result reading is as follows:

USB Wi-Fi result	Reading
Normal browsing	Proves the Windows networking stack, the USB bus, and the comparison network are normal; the problem narrows to the internal-card path, meaning BIOS, PCIe, slot, card, or board level, and the internal-card repair direction becomes clear
USB adapter also abnormal	The problem may lie beyond the internal card: Windows system damage, a USB bus problem, a networking-stack problem, or the control adapter itself may be faulty; try a different USB adapter first to exclude a faulty control tool
USB adapter intermittent	Check the USB port, USB power management, and system stability; do not conclude at once that the internal card is the problem

The USB test has a second value as a temporary workaround. If the internal card needs inspection or parts, the reader can use a USB adapter to keep working instead of stopping the whole machine. The article must state that a USB adapter is a workaround, not a cure, and that long-term USB use must consider USB port occupation, speed, stability, and portability. Note also that the USB test cannot prove 'the internal card is certainly dead'; it can only prove 'the problem is in the internal-card path', which still contains BIOS, slot, power, and board-level possibilities.

Step 6: the teardown or inspection layer, also called the bench or teardown layer. This step applies only to socketed cards (2230 M.2) and must be done by experienced staff under anti-static precautions. The article must clearly separate 'observations the reader can do' from 'operations that must be left to repair staff'. 6A, pre-teardown checks: check the laptop brand service manual for the wireless card position, screw specification, teardown order, and battery disconnection method; power off, disconnect power, remove the battery if removable, and hold the power button for a few seconds to discharge residual power; and photograph the antenna lead positions, screw positions, and card orientation before dismantling. 6B, reseating: loosen the fixing screw, gently remove the card, and check the gold fingers for oxidation, dirt, or damage; check the slot for foreign objects, bent pins, or looseness; reinsert fully and tighten the screw without over-tightening, since an over-tight M.2 screw can crack the card or the board; and check the antenna connectors for looseness, reconnect them, and confirm MAIN and AUX are not swapped. 6C, swap test: swap in a known-good card with a compatible interface generation; if the device returns and stays stable, the old card is indicated; if it still does not appear, the slot, power, BIOS, or board level is indicated; and if the machine fails to boot or BIOS reports an error after the swap, swap the old card back at once, suspecting an allowlist or generation mismatch, and do not force further tries. 6D, slot or board layer: a loose slot, cracked solder joints, abnormal power rails, or PCH-layer anomalies all belong to board-level inspection; board-level inspection needs a multimeter, an oscilloscope, circuit diagrams, and repair experience, and the article may list the inspection items but must not teach readers to measure board power rails themselves, since there is a short-circuit risk. 6E, soldered-down models: if a 1216 soldered-down form is confirmed, there is no card to swap, and diagnosis at this step should stop and be handed to repair staff for board-level or external-workaround assessment; never write 'a soldered module can be swapped by hand', since replacing a soldered module needs hot-air and reballing equipment and technique, not ordinary dismantling.

Step 7: the result summary table. After the steps above, the reader should be able to place their case in one of the following categories:

Category	Evidence	Next step
BIOS disabled	Wireless is Disabled in BIOS, and normal after turning it back on	Keep the BIOS setting and observe
Driver or power management	A driver version or power-setting change clearly improves things	Record the version or setting, keep observing, and rerun the full exclusion on recurrence
Loose card	Returns after reseating and stays stable	Observe, and note whether vibration could loosen it again, as a direction for checking
Faulty card	Normal after a swap test, and the old card also fails in another machine	Replace the card, with a compatible interface generation
Slot or board level	Still absent after a swap, with USB normal	Send for board-level inspection, and assess repair or an external workaround
Soldered-down fault	Soldered form confirmed, software layer fully excluded, USB normal	Send for assessment of board-level work or a long-term external option
Unclassified	All steps tried, with contradictory or unstable results	Bring the full baseline record for inspection, and stop random trying

11. The board power and PCH layer: why board level cannot be guessed at home

If everything above is excluded, the remaining possibility points at the board layer. On power rails, the Wi-Fi card's power comes from a set of board power rails, and the specific voltages and routing vary by model, so the article must not state one voltage value as a universal standard; power ICs, fuses, and ageing or damaged capacitors can all make the card intermittently powered or completely unpowered. On the PCH or SoC, the PCIe root port is integrated into the PCH or SoC, and if the PCH itself is faulty, not only Wi-Fi disappears but neighbouring devices on the same root port may also behave abnormally, such as Bluetooth, a card reader, or other PCIe devices, and that 'abnormal together' pattern is an important clue pointing at board level. On board damage, water exposure, dropping, long-term heat, and static or tool damage during repair can all affect board traces or solder joints, and such damage may not make the card disappear at once but may appear as intermittent disappearance that worsens over time. The diagnostic principle for board-level problems is: the article must not teach readers to do board-level repair themselves; it may state what board-level inspection includes, but the work itself must go to staff with equipment, and that boundary must be written clearly as a safety and responsibility matter.

Any of the following means the article should tell the reader to stop and not try further on their own: a burnt smell, scorch marks, swollen capacitors, or water marks found after dismantling; failure to boot or a BIOS error after a card swap; bent pins, a loose slot, or cracked solder joints; a suspected soldered-down fault; a suspected power-rail or PCH-layer problem; or a machine still within the brand warranty period, where self dismantling may affect warranty — stated as a principle only, with no day count.

12. Hong Kong context: humidity, vibration, and the correct use of comparison networks

Hong Kong humidity is an objective environmental fact, but the article must be careful. It can be written that in a high-humidity environment, the checking priority for gold-finger oxidation and poor slot contact can be raised, and that repair staff will look for oxidation marks when dismantling. It cannot be written that 'because Hong Kong is humid, your Wi-Fi disappeared', since that is a causal leap: oxidation takes time and conditions, does not follow at once from humidity, and many Hong Kong users run machines for years without trouble, so humidity alone cannot explain why this particular machine failed. The correct wording lists humidity as 'one of the items to note during a teardown check', not as 'the cause of the fault'.

Commuting vibration follows the same rule. It can be written that carrying the machine on transport every day and bumping it in a bag can raise the checking priority for a loose card, and that dismantling will check whether screws are loose or the card has shifted. It cannot be written that 'because you ride the MTR every day, the card came loose', unless dismantling actually shows looseness, since otherwise it is only a guess. The correct wording lists vibration as 'one of the questions to ask', such as whether the problem appeared after moving the machine, not as a conclusion.

Home routers, office Wi-Fi, station Wi-Fi, and phone hotspots may only be used as comparison tests, for example: a USB adapter works on the home router and on a phone hotspot, proving the USB adapter and the Windows networking stack are normal; or an internal card is absent on every network, proving the problem is not one particular router but the machine itself. It must never be written that 'because your home router has a problem, Device Manager lost the Wi-Fi card', since a router cannot make a Windows device node disappear. Router problems affect connections, not enumeration. Likewise, it must never be written that station Wi-Fi signal weakness made the card disappear, or that a phone hotspot damaged the card — these are causal errors. The correct logic for comparison testing is: if the card itself is already gone, changing networks will not bring it back; changing networks as a test only fits 'card present but cannot connect', not 'card gone'. If a reader says 'absent at home, present at the office', the distinction must be checked: is it truly 'present' or merely 'present but cannot connect'? The two must be separated; and if enumeration genuinely differs by location, that must be recorded and handed to repair staff, because location itself should not affect PCIe enumeration, and some other variable such as power, temperature, or vibration must sit behind it.

Practical Hong Kong repair notes: before sending a machine for repair, back up data, remember passwords without writing a password on paper for the repairer, using a method the reader controls, and photograph the machine's exterior; quotations must be written, so after a verbal quotation there must be written confirmation before work starts; a written quote after inspection is fair to both sides; and if the repairer says a card must be replaced, the reader has the right to ask the new card model, interface generation, Key, and antenna configuration, and to approve only after confirming all four match.

13. Common misdiagnoses and traps

First, 'it must be a router problem'. Wrong. Router problems affect connections, not device enumeration. If Device Manager has no card, changing routers, restarting routers, changing bands, or changing DNS will not bring the card back. The correct approach is to confirm the Device Manager state first, and only then decide whether the router needs attention. In a card-disappearance case, the router is only a comparison tool, not the culprit.

Second, 'reinstalling Windows will definitely fix it'. Wrong, and high-risk. Reinstalling Windows only resets the software layer, meaning drivers and system settings. If the problem sits in BIOS, the slot, the card, or board level, reinstalling Windows is completely useless, and it also wipes data, requires reinstalling software, and resetting everything, at very high cost. If the problem 'seems fixed' after reinstalling Windows, the likely explanation is only that the reinstall happened to include a full Restart, a re-enumeration, and a different driver version — effects already achievable with driver cleaning and a Restart, without going as far as reinstalling Windows. The article must state plainly: reinstalling Windows is a last resort, not a first step, and before reinstalling, the USB control test must be done to confirm the problem really sits in the system layer. If USB is normal and the internal card is still absent, reinstalling Windows is almost certainly useless.

Third, 'find a registry edit online and it will be fixed'. A high-risk trap that needs prominent warning. Many online tutorials teach registry edits to force the Power Management tab to appear, to force traditional S3 sleep in place of Modern Standby, or to delete UpperFilters or LowerFilters. These changes have several problems: forcing S3 on a Modern Standby machine can cause other sleep, wake, and battery-management problems; a wrong registry edit can make the system unstable or even unbootable; and deleting filters can affect other devices. The article's principle must be: never paste registry commands and tell readers to copy them. If such tutorials must be mentioned, the wording should be that they are high-risk operations, that the registry and a system restore point must be backed up before any change, and that they apply only to specific models; the general reader should not do them and should instead run the non-registry isolation tests first. In Microsoft community answers, Microsoft staff have explicitly warned against reckless registry edits, and the article may cite the spirit of that warning without inventing a specific document number.

Fourth, 'buy a new card and swap it in; the right size is enough'. Wrong, as shown in Sections 9 and 14: the right size does not mean the right Key, the right Key does not mean the right interface generation, and the right interface generation does not mean passing the allowlist. Typical consequences of buying the wrong card are: it fits but is not recognised, leading to the false belief that the new card is also dead; it prevents booting, leading to the false belief that the board is dead; or Wi-Fi returns but Bluetooth is gone, because the Bluetooth USB path has a contact or driver problem during the swap.

Fifth, 'Bluetooth gone means Wi-Fi is dead too', or 'Wi-Fi gone means Bluetooth is dead too'. Wrong, as shown in Section 9: Wi-Fi runs over PCIe and Bluetooth over USB, and they can fail independently. The article must teach readers to record Wi-Fi and Bluetooth status separately, never merged.

Sixth, 'a loose antenna made the card disappear'. Wrong. Antenna problems usually affect signal, not enumeration. Listing the antenna as a disappearance cause sends the diagnosis in the wrong direction. The exception is dismantling damage — tearing off an antenna socket or shorting — which damages the card or board indirectly, and that is dismantling damage, not the antenna's normal failure mode, so the two must be separated.

Seventh, 'run a third-party driver downloader once and it will be fixed'. High-risk: third-party tools can install the wrong version, bundle advertising software, or change system settings; drivers should be downloaded from the laptop brand or card vendor official site, with the model matched character for character. The article must state plainly: do not use driver tools from unknown sources, and if a tool must be used, use Windows Update or the brand's official tool.

Eighth, 'intermittent disappearance means a signal problem'. Wrong. Intermittent disappearance has many forms, as shown in Section 3. A signal problem is characterised by 'card still present, but signal bars drop, speed falls, or connections drop'; a disappearance problem is characterised by 'the card itself is gone'. The two must be separated by the Device Manager state, not by 'sometimes works, sometimes not'.

Ninth, 'do everything at once; if it works, there is no need to know which step worked'. Wrong. Changing BIOS, reinstalling drivers, changing power settings, and swapping the card all at once, and then finding it works, leaves nobody knowing which step was decisive. On recurrence, the whole set must be repeated. The correct approach is one change at a time, with the result recorded. This principle must be repeated throughout the article.

Tenth, other things not to do include: do not repeat a network reset, since a network reset that fails once will not suddenly succeed on the second or third try and only clears settings; do not turn Secure Boot off as a Wi-Fi fix; do not tape over M.2 card disable pins, a forum trick whose principle is to stop the system disabling the card through the disable pin, since it can prevent the system from managing the wireless device normally, affect other devices, or leave the card running when it should not be, and the correct approach is to find out why the system disables the card — BIOS setting, switch, or driver — rather than bypassing system control physically; do not treat a USB Wi-Fi receiver as a diagnostic conclusion, since a USB receiver working does not mean the internal card is fine, and a USB receiver failing does not mean the internal card is dead, as the two are independent paths that must not be inferred from each other; and do not reinstall Windows at once, since reinstalling also clears event logs and driver-install logs and makes later diagnosis harder.

14. Repair, replacement, and upgrade decision framework

Four decision principles apply. First, layer first, quote second: before layered exclusion, any quotation is only an estimate, and the article must teach readers that repair staff need baseline records and control results before they can quote accurately. Second, from cheapest to most expensive, and from reversible to irreversible: BIOS setting, driver or power setting, reseating, card replacement, an external workaround, board-level repair, and replacing the machine, in that order and never reversed. Third, data first: before any dismantling, system reinstall, or board-level work, back up data; Wi-Fi disappearance itself does not delete hard-disk data, but the repair process carries risk, and backup is a mandatory prerequisite. Fourth, do not swap the wrong card for the sake of an upgrade: upgrading a card, for example from Wi-Fi 5 to Wi-Fi 6 or Wi-Fi 6E, must simultaneously satisfy Key, interface generation, allowlist, and antenna conditions, with none missing.

Option	Fits when	Advantages	Disadvantages or risks	Charging wording
Reseating	Loose card, poor contact, oxidised gold fingers	Lowest cost, reversible, immediately verifiable	Needs dismantling; useless if the problem is not looseness	Subject to a written quote after inspection
Replacing the internal card	A swap test confirms the old card is faulty, and the model is socketed	A cure, restores internal Wi-Fi, occupies no USB port	The interface generation must be bought correctly; a wrong purchase will not be recognised; dismantling is needed	Subject to a written quote after inspection; card model, interface generation, and antenna configuration all affect the quote, and the article must not state one fixed price
USB Wi-Fi workaround	Soldered-down fault, waiting for parts, board-level cost too high, short-term cover	No dismantling, plug and play, restores browsing at once	Occupies a USB port, speed and stability depend on the USB adapter, inconvenient to carry, not a cure	Subject to a written quote after inspection, if the repairer supplies the USB adapter
Board-level repair	Slot, power-rail, or PCH-layer problems where a swap still fails	A chance of a cure without replacing the whole machine	Needs professional equipment and experience, takes longer, and some board-level problems may not be worth repairing	Board-level chip work from HK$880; other board-level items subject to a written quote after inspection; note that this price applies only to the confirmed 'board-level chip work' item and must not be applied to other repair items
Replacing the machine	An aged machine, several ageing hardware items at once, board-level cost approaching replacement cost, or a soldered-down plus board-level fault	Solves several ageing problems in one step	Highest cost, and data and settings must be moved	Subject to a written quote after inspection; the replacement decision must compare repair cost with new-machine price, and the article must not directly tell the reader to replace the machine

The confirmed-price usage rule must be followed in writing. The confirmed prices are only these four, and only in their relevant contexts: battery from HK$580; screen from HK$1,280; board-level chip work from HK$880; and keyboard from HK$480. The Wi-Fi card itself, reseating labour, a USB adapter, other board-level items, and machine replacement are all outside the confirmed list and must all be written as subject to a written quote after inspection, with no estimated price, no range, and no 'about how much'. In particular: 'board-level chip work from HK$880' must not be written as 'board repair HK$880', because chip-level work is only one kind of board-level repair and does not mean every board-level problem costs that price; battery, screen, and keyboard prices must not be dragged into a Wi-Fi card paragraph as a comparison, unless that paragraph genuinely discusses whether other problems on the same machine are worth repairing together, since unrelated comparisons only confuse the reader; and every 'from' price must keep the word 'from' and never be written as a fixed price.

The upgrade card purchase checklist requires ten items to be confirmed before buying any internal Wi-Fi card, with none missing: the original card's full model plus hardware ID; the original card interface, standard PCIe plus USB, or CNVio, CNVio2, or CNVio3; whether the target card interface is compatible with the original platform; the CPU or chipset platform, Intel or AMD, and which generation, since CNVio cards fit only specific Intel platforms; the slot Key and size, M.2 2230 A+E or E, and whether it is soldered; the antenna count, how many leads the machine has and how many the target card needs; the BIOS allowlist, whether the official service manual lists supported models; the Windows version, which Windows and driver versions the target card, especially a Wi-Fi 7 card, needs; the driver source, whether the OEM support page carries a driver for the target card or whether a chip-vendor generic driver is needed; and the warranty state, whether the machine is still within warranty, in which case warranty terms must be understood before dismantling. If any of the ten items is unconfirmed, the card must not be ordered. The cost of buying the wrong card is not only money, but also time and dismantling risk.

15. Inspection preparation, recording format, and conclusion

Bringing the following material to inspection makes inspection much faster and avoids making inspectors start from zero: laptop brand plus full model plus BIOS version; Wi-Fi card model plus hardware ID plus driver version, before and after any change; Device Manager state screenshots, both normal and disappeared if they can be captured, including the full error-code text rather than only the yellow mark; the log table, including at least time, trigger, disappearance scope, and recovery method; the list of steps already tried, including the result of each step, not just 'tried everything'; the BIOS wireless setting state and the Secure Boot state; and the Windows Update record, showing whether a driver update preceded the problem.

The suggested log table format is:

Date and time	What was being done before disappearance	Trigger type	Disappearance scope	Device Manager state	Recovery method	Recovery time	Power	Driver version at the time
For example, after waking on a given day	After closing the lid and reopening it	Sleep or lid	Wi-Fi alone, or Wi-Fi plus Bluetooth	Completely absent, grey Code 45, or yellow Code number	Returned by itself, scan, disable and re-enable, Restart, shutdown and start	Minutes or at once	Plugged in or battery	Provider plus version plus date

The conclusion is: disappearance does not mean dead, and disappearance is only a result, while layering is the diagnosis. From BIOS to hidden devices, from drivers and power to the USB control test, and then to the teardown layer, changing one variable at a time and recording the result is what narrows 'it is gone' down to a scope that can be quoted, repaired, and decided. If all steps have been tried and the results remain contradictory or unstable, the honest conclusion is unclassified: bring the full baseline record for inspection, and stop random trying. Pricing and inspection details are on the pricing page at https://www.laptops-king.com/en/pricing and the diagnostic service at https://www.laptops-king.com/en/diagnostic; the inspection location is Room 1209, Daly Plaza, Shek Mun, Shatin, phone +852 2186 8010, WhatsApp +852 6558 6806, hours 12:00–19:00.

Frequently Asked Questions

Q1: Device Manager shows no Wi-Fi card at all. Does that mean the card is dead?

No. No card is only a result: in this boot, Windows did not enumerate the hardware. The cause can sit in BIOS disablement, a power problem, a slot contact problem, an enumeration failure, a faulty card, or a board-level problem, and no single symptom can pick one of them. The correct order is to turn on Show hidden devices and look for a greyed-out item, check Other devices for an unknown device, check the BIOS wireless switch, and then run the USB adapter control test. No driver does not mean no device: if there is not even an unknown device, the driver layer is the less likely suspect, and the BIOS and enumeration layers should come first rather than another driver reinstall.

Q2: How do I tell Wi-Fi dropping apart from a Wi-Fi card disappearing?

Look at Device Manager and ask whether the card is still there. If the card is still listed and the problem is failing to reach a network, frequent disconnections, slow speed, or weak signal, that is a connection problem, and the diagnostic centre of gravity is the router, band, interference, and connection settings. If the card itself is gone — no Wi-Fi switch in Settings, a globe icon in the taskbar, no wireless card in Device Manager, or only a greyed-out hidden item — that is the disappearance problem in this guide, and the centre of gravity is BIOS, enumeration, power, driver loading, and power management. The two problems have different causal chains, and mixing them only sends the diagnosis down the wrong road.

Q3: What does a greyed-out card with Code 45 mean?

Code 45 means Windows considers this hardware device not currently connected to the computer. A greyed-out item means Windows recognised this card before, but did not detect it on this boot — the point is 'present before, absent now', not 'driver never installed'. Possible causes include BIOS disablement, a physical switch turned off, a loose card, a power interruption, an enumeration failure after waking from sleep, or a phantom node left after a driver uninstall. Updating the driver on a greyed-out item is usually useless, because the problem is not the driver version but Windows believing the hardware is absent. Check BIOS and the physical switch first, do a genuine Restart, and then see whether the card turns from grey back to normal.

Q4: What is the difference between Code 10 and Code 43?

Code 10 means the device cannot start: Windows sees the device, but driver loading or device initialisation failed. It is a very general failure signal, and behind it can sit driver corruption, a driver mismatch, a firmware problem, an abnormal power state, or a device start failure of its own. Code 43 means Windows has stopped the device because it reported a problem: the device itself, or its driver, reported an abnormal state, and Windows disabled it to protect the system. Hardware fault carries a little more weight in Code 43 than in Code 10, but neither code can be read as 'the card is dead'. Both are results, not causes, and the handling order is the same in both cases: record the driver version, do a clean install, and then look at the pattern before replacing anything.

Q5: How do Code 28 and Code 31 differ?

Code 28 means the driver for the device is not installed: Windows cannot find a suitable driver for this hardware, and the handling direction is to confirm the model from the hardware ID and then download the matching driver from the laptop brand's official support page for the right model and Windows version. Code 31 means Windows cannot load the driver the device needs: the driver files may exist but be damaged, have a signing problem, or depend on system files that are faulty, and the handling direction is a clean install of the OEM driver followed by reading the driver-install log for the load failure. Do not confuse the two: the first is finding the right driver, the second is clearing a loading obstacle.

Q6: Why does it disappear after sleep or closing the lid, and come back after a Restart?

That pattern points most strongly at the power-management and wake path. Traditional S3 sleep and Modern Standby (S0 low-power idle) have completely different wake paths: on a Modern Standby machine, power is managed centrally by the system, and an individual device may have no Power Management tab at all, which is not a fault. Possible mechanisms include a device power-saving setting failing to re-enumerate on wake, PCIe Link State Power Management failing to re-train the link on wake, or Fast Startup carrying a faulty driver state from the previous session into the next boot. Test one variable at a time: clear the device power-saving setting, set PCIe Link State Power Management temporarily to Off, or temporarily disable Fast Startup, running one sleep-and-wake test after each change and recording the result before changing the next setting.

Q7: I cannot find a Power Management tab. Is the driver broken?

No. On a Modern Standby machine, the Wi-Fi card may have no Power Management tab, because power is handled by the SoC rather than by the traditional per-device power setting. That is a normal design, not something to 'fix' with a registry edit. Tutorials online that teach registry edits to force the tab to appear, or to force traditional S3 sleep in place of Modern Standby, are high-risk operations: a wrong edit can make the system unstable or even unbootable, and the general reader should not do them. During diagnosis, the registry should stay untouched, and the non-registry isolation tests should come first.

Q8: It started disappearing after a Windows Update. Did the update break the machine?

A matching time is only a clue, not a verdict. The real mechanism may be that the driver version pushed by Windows Update is older than the one on the vendor site, or is a generic version that does not fully match the laptop brand's custom settings such as antenna configuration, power limits, or regional rules. It may also be that a new driver demands stricter power-state behaviour than the old firmware can follow, or that the update process left stale versions in the driver store and loading then picks the wrong version at random. The correct approach is to compare the driver version and date with the start time, try Roll Back Driver or a manual install of the vendor's correct version as an isolation test, and then consider hiding the faulty driver update with the official tool. Rolling back without hiding usually fails, because Windows Update is likely to install the same version again soon after.

Q9: A USB Wi-Fi adapter works. Does that prove the internal card is dead?

No. It only proves the problem has narrowed to the internal-card path. The internal Wi-Fi runs over PCIe plus board power plus BIOS enumeration, while a USB adapter runs over the USB bus and completely bypasses the internal card's PCIe path, M.2 slot, and BIOS wireless switch. So a working USB adapter only proves that the Windows networking stack, the USB bus, and the comparison network are normal, while the internal path still contains BIOS, slot, power, and board-level possibilities. Conversely, if the USB adapter also fails, try a different known-good USB adapter first to rule out a faulty control tool, before considering a Windows system or USB bus problem. A USB adapter is also a temporary workaround, not a cure.

Q10: If I buy a new card, is the right size enough?

No. Size is only one of five items that must all be confirmed before buying: the original card model and whether it is socketed or soldered, from the laptop brand's service manual; the Key, A+E or E; the interface, standard PCIe or CNVio, CNVio2, or CNVio3; the 2230 size; and the antenna connector type plus the MAIN and AUX count and line count. Intel tail digits give a first hint: AX200 and AX210 ending in '0' are mostly standard PCIe cards, and AX201 and AX211 ending in '1' are mostly CNVio2 cards, but tail digits are only a hint, and the final check must be the System Interface Type field in Intel ARK or the vendor specification table. A CNVio2 card in a system that only supports standard PCIe will mostly not be recognised, and the result looks exactly like the original fault, which easily leads to the false conclusion that the new card is also dead.

Q11: Wi-Fi is gone but Bluetooth works, or Bluetooth is gone but Wi-Fi works. How do I read that?

Because one two-in-one card runs on two different paths: Wi-Fi usually runs over PCIe, and Bluetooth usually runs over internal USB. Wi-Fi gone with Bluetooth normal points at the PCIe branch, the Wi-Fi driver, or the Wi-Fi part; Wi-Fi normal with Bluetooth gone points at the USB branch, the Bluetooth driver, or USB selective suspend; both gone together points at the whole card losing power, BIOS disablement, a slot problem, or a card fault; both gone but both returning after a Restart leans toward enumeration or the power layer. Always record Wi-Fi and Bluetooth status separately, and never write it as 'Bluetooth gone means Wi-Fi is dead too', or the reverse.

Q12: What should I prepare before inspection? How are charges worked out?

Bring enough material that the inspector can start from the exclusion table instead of from zero: the laptop brand and full model, the Windows version, and the BIOS version; the wireless card model and hardware ID, with driver versions before and after any change; the Device Manager state, including the full text of the error code; the trigger log, recording what happened before each disappearance, whether recovery needed a Restart or a shutdown-and-start, and whether the machine was plugged in or on battery; the list of steps already tried, with the result of each step; and the BIOS wireless setting state. On charges, the only confirmed starting prices are: battery from HK$580, screen from HK$1,280, board-level chip work from HK$880, and keyboard from HK$480. The Wi-Fi card, reseating, card replacement, USB workaround, and other board-level items are all subject to a written quote after inspection. See the pricing page at https://www.laptops-king.com/en/pricing and the diagnostic service at https://www.laptops-king.com/en/diagnostic. Inspection location: Room 1209, Daly Plaza, Shek Mun, Shatin, phone +852 2186 8010, WhatsApp +852 6558 6806, hours 12:00–19:00.

Sources

The sources below support only the mechanism claims and test directions in this article. They do not support conclusions about any individual model, and they do not support any price or success figure. Community sources are used only for reported patterns, never as proof about a machine.

• Microsoft support document: Code 45 explanation, the definition of a device not connected, and the direction of reconnecting the hardware
• Microsoft Q and A: Windows not detecting networking hardware, MediaTek MT7902 case, BIOS wireless switch and driver uninstall plus Restart suggestions
• Microsoft Q and A: Wireless network adapter (code 45), Realtek case, the typical hidden-device and Code 45 picture
• Microsoft Q and A: Power Management Tab Missing, Wi-Fi cards on Modern Standby systems possibly having no Power Management tab, power handled by the SoC, and the warning against reckless registry edits
• Intel official support article: What Are the Intel Integrated Connectivity (CNVi) and Companion RF (CRF) Module?, CNVi and CRF composition, 2230 socketed and 1216 soldered-down forms, CRF usable only in systems specially designed for CNVi
• Intel community: Why does my wireless device not show up?, AX200 and CNVI slot case, AX201 as CNVio2, not backward compatible with CNVio
• Intel community: PCIe vs CNVio and CNVio 2, PCIe plus USB cards versus CNVio2 cards, AX200 and AX210 as PCIe plus USB, AX201, AX211, and AX411 as CNVio2, CNVio3 not backward compatible
• ASRock official FAQ: Which CNVi WiFi modules are compatible with my motherboard?, two kinds of M.2 Key E cards, Intel 300 series supporting CNVio, 400 series supporting CNVio plus CNVio2, 500, 600, and 700 series supporting CNVio plus CNVio2
• ASRock M.2 (Key E) Socket Support List, System Interface Type comparison across models
• Microsoft Q and A: Wireless network adapter Error code 43 and Code 10, Qualcomm Atheros QCA9377 case, Code 10, 43, and 45 appearing in turn
• Microsoft Q and A: m.2 WiFi adapter not visible in Windows 10 Device Manager, but works in Linux, AX210 case, a device seen in Linux but not in Windows, pointing at the Windows driver or enumeration layer
• Third-party tutorial, as phenomenon evidence only and not as an official source: WiFi network adapter isn't waking up or goes missing after Sleep in Windows 11, Modern Standby versus traditional sleep, Fast Startup residual state, and the PCIe Link State Power Management role in post-wake disappearance

Unverified items, which must not be written as conclusions: the specific titles and wording of Realtek, MediaTek, and Qualcomm official support pages; the specific chapter titles of Microsoft's official Modern Standby technical document; the document number of Microsoft's official Device Manager error-code list; the official specification of I-PEX MHF4 connectors; the specific names of BIOS options by brand; any causal relationship between Hong Kong humidity or vibration and Wi-Fi disappearance; any fault proportion, average repair time, or success-rate figure; any definite statement about CNVio3 backward compatibility; and any universal standard value for board power-rail voltage.