GPU Not Detected or Slower After a Motherboard BIOS Update? Reset Defaults, UEFI/CSM, ReBAR, Above 4G & PCIe Generation Diagnosis Guide

| HK GPU Fix

GPU Not Detected or Slower After a Motherboard BIOS Update? Reset Defaults, UEFI/CSM, ReBAR, Above 4G & PCIe Generation Diagnosis Guide

Key Takeaways

After a motherboard BIOS update, a missing or slower graphics card is often a settings regression, not a failed card: the update can restore defaults, so compare each setting methodically before assuming hardware failure and record every change so you can isolate the single setting responsible for the change in behaviour.

What This Article Covers, and What It Does Not

This article covers exactly one trigger event: a computer whose graphics card worked normally before a motherboard BIOS update, and then, after that update, is not detected, produces no display, is locked to a low resolution, or performs noticeably worse under load. The protagonist throughout is the change introduced by that update, and in particular the settings the update may have restored to factory defaults. Everything that follows is organised around distinguishing a settings regression from other explanations that happen to coincide in time.

Three neighbouring topics are deliberately excluded, and keeping them separate matters. First, this is not a general guide to installing a new graphics card in an older computer. In that situation the starting point is a first-time hardware combination, and the questions are about power delivery, slot compatibility, and the support range of the older motherboard. Second, this is not a guide to a black screen after installing a new graphics driver. In that situation the starting point is a change inside the operating system, and the diagnostic path runs through driver versions, driver removal, and operating-system behaviour. Third, this is not a ReBAR-only guide. Resizable BAR is only one of several settings that an update can restore, alongside boot mode, Compatibility Support Module behaviour, Secure Boot, Above 4G Decoding, PCIe generation, slot configuration, display-output priority, and saved profiles. Treating any one of them as the whole answer risks missing the setting that actually changed in your machine.

Why does the distinction matter in practice? Because the remedy follows the cause. If a settings regression is treated as a hardware failure, a working card may be replaced unnecessarily, and the replacement may inherit exactly the same restored defaults and show exactly the same symptom. If a genuine hardware fault is treated as a settings problem, hours can be spent changing firmware options that cannot compensate for a physical fault. The diagnostic order below exists to prevent both errors: establish what changed, locate the layer at which detection stops, compare settings one at a time, and only then judge which explanation fits the evidence.

Why a BIOS Update Can Affect a Graphics Card

The motherboard BIOS, in modern machines usually called UEFI firmware, is the first program to take control of hardware when a computer powers on. It initialises the processor, memory, storage devices, and expansion slots, including the PCIe slot that holds a discrete graphics card. For a display to appear, the card must be initialised correctly during this early phase, and control must then be handed to the operating system and its driver in an orderly way. Any change to how that early phase is configured can therefore change whether a display appears at all, whether the card is recognised, and how well it performs once the system is running.

A BIOS update frequently restores some settings to factory defaults. Sometimes the update notes say so explicitly. Sometimes the restoration happens because the new firmware uses a different settings format, because a security-related option has been redesigned, or because the update process deliberately clears settings that cannot be carried forward safely. The result is the same from the user's point of view: options that were deliberately changed to suit a particular graphics card, a particular operating-system installation, or a particular boot arrangement may quietly revert, without any warning appearing on screen during normal use.

A second mechanism is the loss of saved profiles. Many motherboards allow a complete set of settings to be saved under a profile name so that it can be restored later. After an update, an older profile may no longer be compatible with the new firmware, or the profile list may be cleared. A user who relied entirely on a saved profile, without a written record of the individual settings inside it, then has no reliable baseline. They remember that the machine worked well before, but they cannot say, setting by setting, what the previous configuration actually was. That gap is not a minor inconvenience. It removes the very comparison that makes diagnosis possible.

A third point deserves emphasis: a factory default is not the same as a mistake. Defaults are chosen so that the widest possible range of hardware combinations can boot. They are not necessarily tuned for a specific card, a specific operating-system installation method, or a specific display arrangement. A default boot mode may not match how the operating system was installed. A default PCIe setting may negotiate differently from a manually chosen one. A default display priority may favour integrated graphics over a discrete card. Before the update, a manual change hid each of these mismatches. After the update, the mismatch is exposed, and it looks like a new fault even though the hardware has not changed at all.

Classify the Symptom Before Changing Anything

Symptoms after a BIOS update fall into three broad groups, and the group determines where to look first. The first group is no display at all. The screen reports no signal, the motherboard debug indicator may stop at the VGA or boot stage, fans spin, but the system never reaches the operating system. This group most often points to initialisation or boot-mode issues: the wrong display output has priority, the boot mode does not match the installed system, or slot initialisation is configured in a way that prevents the card from starting correctly.

The second group is detection with degraded behaviour. The operating system recognises the card, but the resolution is locked low, the refresh rate cannot be raised, frame rates in games or graphics workloads are clearly below their pre-update level, or the system becomes unstable under graphics load. This group points more often toward PCIe generation, the Above 4G Decoding and Resizable BAR combination, driver interaction, or a mismatch between firmware settings and the operating system's expectations. The card is present, but the conditions under which it operates have changed.

The third group is intermittent behaviour. Sometimes the machine boots with a display and sometimes it does not. A cold boot fails while a warm restart succeeds, or the reverse. Intermittent symptoms are the easiest to misjudge, because they invite explanations based on loose connections or power supply behaviour. Firmware settings can also produce intermittency, especially when boot mode, Secure Boot, and PCIe settings sit at a boundary where initialisation timing and ordering become sensitive. Intermittent cases are precisely the ones where changing one setting at a time, and recording each result, matters most. Without that discipline, every restart feels like a new random event rather than evidence.

Classification is not labelling for its own sake. It is triage. No display means start with boot and display-output layers. Degraded performance with detection means start with PCIe and related feature settings. Intermittency means slow down, change one variable at a time, and treat each boot result as a data point rather than as a verdict.

The Diagnostic Order, Step by Step

Step 1: Establish what changed. Write down the BIOS version before the update and the version after it, if both are known, together with the date of the update, how the machine behaved before it, and when the symptom first appeared. Then ask three questions and answer them in writing. Was the machine fully normal immediately before the update? Did the symptom appear immediately after the update, or after a delay? Was anything else changed at the same time, such as a driver update, an operating-system update, moving the machine, cleaning inside the case, reseating hardware, or changing a cable or monitor? If several changes happened together, they must be separated. An assumption that the BIOS update caused the problem, made without checking for simultaneous changes, is not a diagnosis. It is a guess that happens to be convenient.

Step 2: Check the layers of detection. Detection happens in layers, and each layer corresponds to a different stage of startup. The first layer is power-on self-test behaviour: whether the screen receives a signal, where a motherboard debug indicator stops, whether keyboard indicators respond, and whether fans and storage activity look normal. The second layer is the firmware device list: whether the BIOS or UEFI setup lists the discrete card, what it reports for the slot, and which device is selected as the first display output. The third layer is the operating system: whether Device Manager or its equivalent lists the card, whether any error indication is present, and what resolution and refresh rate are available. The fourth layer is the driver: whether the driver version matches the pre-update version, and whether the driver recognises the full model and memory configuration of the card.

Writing down the result at each layer reveals where the chain breaks. If the firmware layer already fails to list the card, the problem is likely at initialisation, slot configuration, or boot-mode level. If firmware lists the card but the operating system does not, attention shifts to the operating-system and driver layers. If the operating system lists the card but performance has changed, attention shifts to PCIe generation and related feature settings. Skipping this layered check and jumping straight to changing settings is a common source of wasted effort, because it treats every layer as equally likely when the evidence already narrows the field.

Step 3: Compare BIOS settings one by one, changing only one setting at a time. List every firmware setting related to graphics: boot mode, Compatibility Support Module state, Secure Boot state, first display device or display-output priority, Above 4G Decoding, Resizable BAR, PCIe slot generation, and slot bifurcation or slot configuration where present. For each setting, record its current value, the value you change it to, the boot result, and the change in symptoms. Only then move to the next setting.

The one-change-at-a-time rule is not pedantry. It is the only way to establish causation. If five settings are changed together and the problem resolves, no one knows which change mattered, and the next update may undo the unknown critical change again. If five settings are changed together and the problem worsens, no one knows which change caused the worsening. A slower, recorded sequence produces a reusable result: a known stable combination, and a written record of which setting was decisive. That record is also the most useful thing to hand to a technician if inspection later becomes necessary.

Step 4: Separate four explanations. The first explanation is a settings regression: a symptom maps clearly onto a specific setting change, restoring that setting resolves the symptom, and the effect can be reproduced by toggling the setting. The second explanation is a failed or corrupted update: settings do not save, the boot process itself is repeatedly unstable, the firmware version display is inconsistent, or abnormalities appear that have no relationship to graphics settings. The third explanation is a coincidental hardware fault: the update merely coincides in time with a physical change, such as a connection, cable, power-delivery, or card condition that changed around the same period. The fourth explanation is a simultaneous driver or operating-system change: a driver update or system update performed on the same day changed resolution, refresh rate, stability, or performance, and is mistaken for a firmware effect.

Separation relies on reproducibility and correspondence. A settings regression shows a stable correspondence between one setting and one symptom. A driver or system change can be separated by comparing driver versions and system update history against the timeline. A coincidental hardware fault typically shows no stable correspondence with firmware settings: changing firmware options does not produce a repeatable improvement. A failed update typically shows instability beyond graphics settings, such as settings that will not persist or a boot process that fails in varying ways. Keeping these four explanations distinct prevents the most common diagnostic error, which is to settle on the first plausible cause and stop testing the others.

Step 5: Decide when to stop. Diagnosis should not continue indefinitely. Stop and seek inspection when any of the following applies: repeated boot failures persist after the key settings have been compared one by one; firmware settings revert on their own after being saved; even a basic display configuration with minimal variables produces no display; a debug indicator persistently stops at the same stage; or you can no longer state with confidence which settings you have changed. Continuing to change options past this point risks turning a clear settings problem into a confused state that is harder for anyone, including a technician, to untangle. A clean stopping point, supported by written notes, is a successful outcome for a diagnostic session, not a failure.

Settings Comparison Table: What an Update Most Often Restores

SettingChange an update can introduceSymptom it can produceHow to check it
Boot mode, UEFI or CSMReverts from UEFI to CSM, or from CSM to UEFINo display, failure to find the boot device, boot process stopping unexpectedlyConfirm that boot mode matches how the operating system was installed, and record both values before changing anything
Secure BootEnabled, disabled, or key state changedBoot blocked, a particular boot method refusedRecord the before-and-after state, and change it only one step at a time with a note of the result
First display deviceReverts from discrete graphics to integrated graphics priorityScreen shows no signal even though the machine itself has bootedConfirm which output the monitor cable is attached to, then confirm the first display setting matches it
Above 4G DecodingReverts from enabled to disabledLarge-memory card detection anomalies, Resizable BAR unable to work as beforeAlways check this together with Resizable BAR, never in isolation
Resizable BARReverts from enabled to disabledPerformance in some workloads falls back from the pre-update level, while the card remains detectedCompare its state with Above 4G Decoding as a pair, and note the combination, not just one value
PCIe slot generationReverts from a manual generation to automatic negotiation, or automatic negotiation produces a different resultSlower performance, unstable transfers, intermittent detectionNote the negotiated generation reported by firmware or system tools, and compare it step by step
Saved profilesProfile list cleared or older profiles become incompatibleSeveral settings revert at once, making step-by-step tracing difficultKeep a written list of key settings, and do not rely on a saved profile as the only record

This table is a checklist, not an instruction to change everything at once. Start with the row most closely related to your symptom, verify it, record the result, and only then move to the next row.

Each Setting in Detail

UEFI and CSM represent fundamentally different boot approaches. UEFI is the newer boot method, while the Compatibility Support Module provides compatibility with older, legacy-style booting. An operating system installed in one mode generally expects to boot in that same mode. If an update restores the opposite mode, the machine may fail to find its boot device, or display initialisation may diverge from the expected sequence. Before changing this setting, establish which mode your installation originally used. Choosing the newer-sounding option on principle, without that baseline, can create a new problem while trying to solve the old one. The correct question is not which mode is newer, but which mode matches the existing installation.

Secure Boot is a boot-security mechanism that checks whether boot software carries a recognised signature. An update can change its enabled state or its key state, and a previously bootable combination can then be refused. This setting should not be toggled casually, because it affects the whole boot chain, not only graphics. Record its post-update state first, compare it with whatever pre-update state you can establish, and change it one step at a time while noting the boot result. If Secure Boot is not the decisive setting, leaving it in an undocumented state only adds confusion to the next diagnostic round.

Display-output priority is the easiest setting to overlook and the easiest cause of a false no-display verdict. Many motherboards default to integrated graphics output when integrated graphics are available. If the monitor cable is attached to the discrete card, but firmware prioritises integrated output, the screen will show no signal while the machine itself may have booted normally. Before entering detailed firmware comparisons, confirm where the cable is attached and what the first display device setting selects. This simple check can eliminate a whole category of apparent failures, and it costs far less time than reseating hardware or reinstalling drivers.

Above 4G Decoding and Resizable BAR must be examined as a pair. Above 4G Decoding allows the system to assign a larger address space to large devices, and Resizable BAR allows the processor to access the full memory of the graphics card directly. In many configurations they are intended to work together. If an update restores one and not the other, the combination becomes inconsistent: the system may still detect the card, but a feature that previously worked may no longer engage, or performance may fall back in workloads that previously benefited. Checking only one of the pair gives an incomplete picture. Always record both values together, as a combination, and test changes to the combination as a single documented step.

PCIe generation determines the transfer specification negotiated between the slot and the card. Automatic negotiation is supposed to select the highest specification supported by both sides, but in real combinations the negotiated result can vary with firmware version, slot settings, risers or adapters, and cable or connection conditions. After an update, automatic negotiation may behave differently from before, and the result may be a lower generation than the pre-update configuration used. The effect is often a performance reduction rather than a complete detection failure, because the card is still present, only operating under different transfer conditions. Firmware or system tools that report the negotiated link speed and width are useful here, because they turn a vague impression of slowness into a comparable figure that can be checked before and after each change.

Slot selection and slot bifurcation also deserve attention on boards with more than one PCIe slot. Slots are not always equal: they may offer different bandwidth, different generation support, or different sharing arrangements with other devices. An update may restore a default bifurcation or slot priority that does not match the slot actually in use. If a card was placed in a secondary slot for a specific reason, the post-update defaults may no longer reflect that arrangement. Verification should therefore start from physical reality, which slot the card occupies, and then compare the firmware settings for that specific slot, rather than assuming that all slots behave identically.

The loss of saved profiles is a process problem, not a single-setting problem. Users who never wrote down their key settings discover the loss only after an update clears the profile list. The remedy is not to guess the old values, but to build a new written baseline from the current, stable configuration once it is found. That baseline should list boot mode, CSM state, Secure Boot state, first display device, Above 4G Decoding, Resizable BAR, PCIe generation, and slot position. Written before the next update, it turns the next comparison into a routine check instead of a reconstruction from memory. A profile stored in firmware is convenient, but a written list is the only record that survives a firmware format change.

Boundary Cases: Not Every Post-Update Problem Is a Settings Regression

The first boundary case is a problem with the update itself. An interruption during updating, an incomplete update file, or an abnormal firmware state after updating can leave the board unstable in ways that go beyond graphics settings. The telltale signs are problems that are not confined to graphics: settings that will not save, a boot process that fails in inconsistent ways, a firmware interface that behaves abnormally, or several unrelated settings showing anomalies at once. In this situation, continuing to adjust graphics settings is the wrong response. The firmware layer itself needs careful handling, and the correct step is to stop and seek inspection rather than to keep toggling options.

The second boundary case is a coincidental hardware change. Hardware does not fail because firmware was updated, but the timing can mislead. Around the same period, a machine may have been moved, cleaned, opened, or reconnected, and any physical change can affect contact, cable seating, or power delivery. The distinguishing test is reproducibility: a settings regression shows a stable, repeatable correspondence between one setting and one symptom, while a coincidental physical fault typically shows no such stable correspondence with firmware options. If changing firmware settings produces no repeatable improvement, physical-layer checks belong back on the list.

The third boundary case is a simultaneous driver or operating-system change. Many users update firmware on the same day that they update a driver or install system updates. Driver changes can affect resolution, refresh rate, graphics performance, and stability, and system updates can change how a driver loads. Without a record of driver versions and system update dates, the two timelines blur into one event, and the firmware update takes the blame by default. The remedy is to keep driver version, system update history, and firmware version in the same diagnostic note, and to compare all three timelines before assigning cause.

The fourth boundary case is the monitor and cable layer. A monitor set to the wrong input, a loosened cable, or a changed adapter can produce no display or an abnormal resolution that looks exactly like a firmware settings problem. Checking monitor input source and cable attachment before entering firmware comparisons avoids a long detour through settings that were never the cause. The order matters: physical display path first, then firmware settings, then operating-system and driver layers, unless the layered detection check already points elsewhere.

The fifth boundary case is multiple settings reverting at once. This is the hardest pattern, because the symptom is the sum of several changes rather than the effect of one. Boot mode and display priority reverting together can produce no display. Resizable BAR and PCIe generation reverting together can produce a performance fallback. The method remains one change at a time, but expectations must adjust: the first change may bring only partial improvement, not full recovery. Recording partial improvements is essential, because the full picture emerges only when the steps are accumulated. A diagnostic note that records only complete success or complete failure will discard the partial evidence that actually explains a multi-setting regression.

Common Misconceptions

Misconception one: a slower result after an update means the new firmware has made the card worse. Firmware does not directly change the hardware capability of the card. It changes how the motherboard initialises and configures the card. If the pre-update settings combination is restored, performance often returns to its pre-update level. Blaming the firmware version itself, without comparing settings, misses the adjustable cause and leaves the machine in a degraded state that could have been corrected.

Misconception two: not detected means the card has failed. Detection has layers, and failure at each layer means something different. Firmware-level non-detection, operating-system-level non-detection, and driver-level non-recognition point to different stages. Many settings regressions cause non-detection only at the initialisation or priority stage, not because of hardware damage. Judging hardware failure before the detection layers have been separated is one of the most common misjudgements in post-update troubleshooting.

Misconception three: restoring all defaults and reconfiguring from scratch counts as diagnosis. Restoring defaults is a starting point, not a diagnosis. Without a record of each change, the next update will reproduce the same confusion, because no one will know which setting was decisive. The value of diagnosis lies in the record, not in the number of settings changed. A short, accurate note beats a long, undocumented sequence of trial changes.

Misconception four: enabling Resizable BAR always improves performance, and disabling it always explains a slowdown. The effect of Resizable BAR depends on the whole combination, including Above 4G Decoding, driver behaviour, the application or workload, and the card itself. Attributing every performance change to Resizable BAR alone ignores PCIe generation, boot mode, and other settings that may have changed at the same time. Resizable BAR should be treated as one item on the checklist, not as the sole answer.

Misconception five: without a pre-update record, comparison is impossible. Even without a complete old record, diagnosis can start from the present symptom and the present detection layers, testing one setting at a time and building a new baseline as it goes. The old combination may not be recoverable in full, but a new stable combination can be established and written down. That new baseline then serves as the comparison point for the next update, which is the practical goal. Waiting for a perfect old record, while the machine remains unusable, helps no one.

What to Record Before Seeking Inspection

If the stopping point is reached, the following notes will make inspection far more efficient. First, the motherboard model and the firmware versions before and after the update, if known. Second, the graphics card model and the slot it occupies. Third, pre-update behaviour and post-update symptoms, stated precisely: no display, not detected, or degraded performance, and under what conditions. Fourth, the layered detection results: whether firmware lists the card, whether the operating system lists it, and which driver version is installed. Fifth, every setting already tried, with its value before the change, its value after the change, and the result of each change. Sixth, any simultaneous changes: driver updates, system updates, moving, cleaning, or cable changes. Seventh, monitor model, cable type, and monitor input source.

Write these notes down as text. Spoken descriptions easily omit decisive details, such as which setting was changed and what happened after that specific change. A written note allows a technician to continue from the point where you stopped, instead of repeating steps you have already tested. It also protects you from a second round of undocumented changes that could obscure the original pattern.

Frequently Asked Questions

Q: After a BIOS update there is no display at all. What should I check first?

A: Confirm the monitor input source and cable attachment first, then note where any motherboard debug indicator stops. Next, enter firmware setup and check first display device and boot mode against the combination your system originally used. Do not assume hardware failure at this stage, and do not change several settings at once.

Q: Why would Resizable BAR and Above 4G Decoding change after an update if I did not touch them?

A: Because an update can restore settings to factory defaults or invalidate an older settings format. These two settings work as a pair, so restoring one without the other creates an inconsistent combination. Record both values together and compare them as a pair, rather than checking one in isolation.

Q: Can a PCIe generation change cause a card to disappear completely?

A: A generation change alone is less likely to cause complete non-detection. It more commonly causes slower performance or intermittent instability, while the card remains detected. Complete non-detection should first prompt checks of initialisation, slot configuration, display priority, and boot mode, before generation is treated as the primary cause.

Q: Is it acceptable to restore all of my remembered settings at once?

A: If you are certain of every old value, restoring them together can work, but each change should still be recorded. If you only roughly remember the old values, changing one setting at a time is safer, because it reveals which change brings improvement and which brings deterioration. Certainty, not speed, should decide the approach.

Q: How do I distinguish a settings regression from a driver problem?

A: Look at detection layers and timelines together. If firmware already fails to list the card, or if boot mode and display priority have clearly changed, a settings regression is more likely. If firmware and the operating system both list the card, but the driver version changed on the same day, a driver-layer explanation deserves equal attention. Compare firmware version, driver version, and system update dates in one note before assigning cause.

Q: I have compared all the key settings and the problem remains. Should I seek inspection?

A: Yes, if you have compared boot mode, CSM state, Secure Boot, display priority, Above 4G Decoding, Resizable BAR, and PCIe generation one by one, separated the detection layers, and excluded simultaneous driver and system changes, yet the problem persists, or if boot failures repeat or settings will not save, then stop and seek inspection. Further undocumented changes will only make the original pattern harder to recover.

Conclusion: The Update Is Not the Enemy, the Missing Record Is

A BIOS update is a normal part of maintenance. The difficulty usually lies not in the update itself, but in the absence of a comparison after it. A written list of key settings, made before an update, compared one by one after it, with each change and each result recorded, turns the next update from a surprise into a routine check. That habit is the real protection, more than any single setting value.

If you have reached the stopping point, or if you are no longer certain which settings you have changed, bring the diagnostic notes described above when seeking inspection. They are far more useful than a general statement that the card is not detected. HK GPU Fix is located at Dali Plaza, Room 1209, Shek Mun, with telephone 2186 8010 and WhatsApp 6558 6806.

Related articles

Related services

If you need Display output repair, see these service pages: