GPU-Z Can't Read Your Graphics Card or Shows Unknown? Device Manager, Version, Driver & VBIOS Diagnosis

Key Takeaways
GPU-Z showing Unknown, a blank field, or all zeroes does not by itself mean the graphics card is broken. GPU-Z is a reader, not a measurer: it reports what the VBIOS, the driver, and Windows tell it. This guide follows four decision points and seven steps to find which link in that reading chain has broken.
What This Problem Is — and What It Is Not
Picture the scene: a graphics card is installed, the computer starts, and there is a picture on the screen. Then GPU-Z is opened, and something is wrong. The name field says Unknown. Or the BIOS Version field is blank. Or every sensor reading is 0, or stuck at a default value. Or the card's name is there, but the memory size, the clock speeds, or the driver details do not match what the card's specification page says.
That is the problem this guide covers, and it is worth defining precisely, because the first impression it creates — "the card must be broken" — is usually wrong. GPU-Z is a reader, not a measurer. It does not open the card and count the memory chips, or probe the graphics processor directly to find out which model it is. It reads three things: the information written in the card's VBIOS, such as the model name, the memory capacity, and the BIOS version; the information reported by the driver, such as the driver version, the WDDM model, and the DirectX feature level; and the live readings reported by the sensors, such as temperature, clock speed, power draw, and fan speed.
So "GPU-Z cannot read the card" does not mean "the card is broken". It means "one link in the reading chain has broken". The reading chain has four links: enumeration — whether Windows recognises the card at all; the driver — whether the manufacturer's driver is installed and working; the version — whether this copy of GPU-Z recognises the card's device ID; and the VBIOS — whether the BIOS can be read, and what it says. Each link fails in a different way, and GPU-Z looks different depending on which link has failed. That difference is the whole diagnosis.
Four conditions must all be true before the steps below apply:
- The symptom is in GPU-Z itself: a name shown as Unknown, a blank or garbled field, readings that are all 0 or stuck at default values, or a model name that does not match the card's official specification.
- The computer is a normal desktop computer, with the card installed in it and a display connected to it. Virtual machine passthrough and headless operation without a display are separate situations with their own limits, and they are covered in the boundaries section below.
- There is no driver error code showing in Device Manager. A card with a yellow warning and an error code such as Code 43 follows the error-code diagnosis, not this guide.
- The question is about the card's identity — its name, its device ID, its BIOS version — or about GPU-Z reading nothing at all. A card whose identity reads correctly but whose temperature or fan speed readings look wrong is a sensor problem, not this problem, and it follows the sensor diagnosis.
Why does the "reader, not a measurer" frame matter so much? Because it changes what a blank field means. A measurer that finds nothing has found that nothing is there. A reader that finds nothing has found that one of its sources is silent. The card can be working — producing a picture, running a game — while a reader cannot describe it, because the reader's information comes from elsewhere. A computer with a picture on the screen already proves that the card has power and that its output works. The question is not "is the card alive", but "which link in the reading chain is silent, and why".
A note on the structure of this guide: the four decision points, the seven steps, and the six misconceptions are the whole structure. Every section below refers back to them, and the numbers stay the same throughout — four decision points, seven steps, six misconceptions — so that a result at any step can be traced back to the decision point it tests. The four decision points are also the order of the diagnosis. Each one says how to tell whether the break is at that point, and where to go next. None of the four is expanded into a full guide of its own here; each neighbouring subject — the driver layer, the enumeration layer, the VBIOS layer, and the version layer — has its own guide, and this guide takes only the one sentence it needs from each.
The Four Decision Points
Whether GPU-Z can describe a graphics card is decided by four decision points working together. None of them acts alone, and a failure at any one of them can produce a blank field, an Unknown, or a row of zeroes. Understanding all four before touching anything is what keeps the diagnosis from going in circles — and from replacing a card that was never the problem.
Decision Point One: Device Manager — Is the Card There, and What Is It Called?
The first decision point is Windows Device Manager, under Display adapters. It gives one of four answers, and each answer sends the diagnosis down a different road:
- The card is listed, its name is correct, and there is no warning. Enumeration is normal. The problem is in GPU-Z itself, or in the driver reporting layer, and the diagnosis moves to Decision Point Two.
- The card is listed, but its name is "Microsoft Basic Display Adapter" or "Standard VGA Graphics Adapter". That is the generic driver state: Windows is using its own basic driver because the manufacturer's driver is not installed or not working. The diagnosis moves to Decision Point Three.
- The card is listed with a yellow warning mark. That is a driver-layer problem, and it is divided by error code. A Code 43 has its own guide; other error codes have theirs. This guide covers the case with no error code, so a yellow warning is a boundary, not a starting point.
- There is no card listed at all. That is an enumeration problem — Windows does not see the card — and it follows the card-not-recognised diagnosis, not this guide.
Device Manager is the first decision point because it is the cheapest and most informative test available, and because it is independent of GPU-Z. If Device Manager cannot see the card, no version of GPU-Z will see it either, and updating GPU-Z or reinstalling a driver is wasted effort until enumeration is fixed. If Device Manager sees the card correctly, GPU-Z's failure is already narrowed to the later links. Skipping Device Manager and going straight to reinstalling a driver is the most common way this diagnosis goes wrong: it changes a layer that may be perfectly sound, and it destroys the evidence about which layer was actually broken.
Decision Point Two: GPU-Z Version — Does This Copy of GPU-Z Know This Card?
GPU-Z carries a database of graphics cards, keyed by device ID. When a new card is released, older versions of GPU-Z do not yet have that card's device ID in their database, and they cannot name it. The result is exactly the symptom this guide covers: Unknown, or a blank name, on a card that is working normally.
The test is simple: open GPU-Z, note its version number, and compare it with the current official version on the TechPowerUp website. If the installed version is older than the card's release year, the failure to recognise the card is expected, not surprising. Update GPU-Z to the current version and try again. This step costs nothing, requires no reinstallation, and carries no risk, which is why it comes before any driver work.
If GPU-Z is already the current version and still cannot recognise the card, the version layer is cleared, and the diagnosis moves to Decision Point Three or Four. A current version that cannot name a card is itself informative: it says the problem is not that the database is out of date, but that something further down the chain — the driver, or the VBIOS — is not supplying the information GPU-Z needs.
One caution belongs here. GPU-Z does not install or update graphics drivers. Updating GPU-Z and updating the driver are two different actions, and they must not be confused. Updating GPU-Z changes only the reader. Updating the driver changes one of the sources the reader reads from. The two are tested separately, one at a time, so that the result of each test can be attributed to the change that produced it.
Decision Point Three: Driver State — Generic, Manufacturer, or Damaged?
The third decision point is the state of the driver, read in Device Manager under the card's Properties and the Driver tab. The provider name there is the evidence: NVIDIA, AMD, or Intel means the manufacturer's driver is installed; Microsoft means the generic driver is in use.
- Generic driver, shown as "Microsoft Basic Display Adapter": install the manufacturer's driver, then open GPU-Z again and compare. The generic driver reports very little about the card, and GPU-Z cannot describe what the driver does not report.
- Manufacturer's driver installed, but GPU-Z shows all readings as 0 or at default values, with only the name present: the driver is damaged, or the communication between the driver and GPU-Z has broken. A clean driver reinstall is the comparison test for this state. The full clean-reinstall procedure, including safe mode and a driver cleanup tool, has its own guide; this guide needs only the one sentence that the comparison exists and when to run it.
- Manufacturer's driver installed, GPU-Z reads the name but cannot read the sensors: that is a sensor-layer problem, not an identity problem, and it follows the temperature and fan speed guides.
The driver state matters because GPU-Z reads through the driver. A card can be producing a picture while its driver reports almost nothing — the generic driver is designed to produce a basic picture and little else. That is why "there is a picture, so the driver must be fine" is a mistake. A picture proves the output works. It does not prove the driver is reporting the card's identity, its memory, or its sensors.
Decision Point Four: VBIOS and Device ID — Can the BIOS Be Read, and What Does It Say?
The fourth decision point is the VBIOS, read in GPU-Z through two fields: "BIOS Version" and "Device ID". Those two fields, compared with the card manufacturer's official specification page, give three possible answers:
- The BIOS Version field is blank or garbled. The VBIOS cannot be read. That is a VBIOS-layer problem, and it follows the VBIOS diagnosis. The full recovery procedure for a damaged or wrongly flashed BIOS has its own guide; this guide needs only the recognition that a blank BIOS Version field points there.
- The device ID is unrecognised, or the model name does not match the official specification. That raises the possibility of a card whose BIOS has been modified to report a false model name. GPU-Z reads the number the VBIOS writes down; it does not measure the hardware directly, so a modified VBIOS makes GPU-Z report the wrong name faithfully. The model name must be checked against the official specification page and the device ID, never trusted on its own.
- The card has a dual BIOS switch. Try switching to the other BIOS and reading again. A card with two BIOS chips can be read from either side, and one side may be readable when the other is not. The differences between silent and performance BIOS positions have their own guide; this guide needs only the one sentence that the switch is a comparison test.
Taken together, the four decision points give the diagnosis its shape. Device Manager tells whether Windows sees the card at all. The GPU-Z version tells whether the reader knows this card. The driver state tells whether the manufacturer's driver is reporting. And the VBIOS tells whether the card's own identity can be read. Each of the seven steps below tests one of these points, in an order that clears the cheapest and most informative tests first.
The Seven Diagnostic Steps
Work through these steps in order, and change only one thing at a time. If GPU-Z is updated, the driver is reinstalled, and the BIOS is switched all at once, and GPU-Z suddenly reads the card, there will be no way to know which change did it — and no way to know which one to check first if the problem returns. One variable at a time, and GPU-Z reopened after each change.
Step 1: Look at Device Manager First
Open Device Manager, expand Display adapters, and write down what the card is called and whether there is a yellow warning mark. This step decides which road the rest of the diagnosis takes, and it must not be skipped in favour of reinstalling a driver directly.
The four possible answers are the four branches of Decision Point One: correct name with no warning; generic name such as "Microsoft Basic Display Adapter"; a yellow warning; or no card at all. Record the exact name shown, word for word, including whether it says Microsoft or the manufacturer's name. That exact name is the evidence, and a paraphrase of it is not. A note that says "Device Manager shows Microsoft Basic Display Adapter" is useful; a note that says "Device Manager looked wrong" is not.
If there is no card listed at all, leave this guide: the problem is enumeration, and the card-not-recognised diagnosis is the right path. If there is a yellow warning with an error code, leave this guide for the error-code diagnosis. The steps below are for the case where Device Manager shows a card — correctly named or generically named — with no error code.
Step 2: Compare With a Second Tool
Open a second tool that reports the graphics card independently of GPU-Z: Task Manager, under Performance and then GPU, or the DirectX Diagnostic Tool, known as dxdiag, under its Display tab. Write down whether the second tool recognises the card.
If the second tool also cannot recognise the card, the problem is in the enumeration or driver layer, not in GPU-Z. Two independent readers failing together points at the source they share, not at either reader. If the second tool recognises the card and only GPU-Z cannot, the problem is locked to GPU-Z itself — its version, or its database — and the driver and enumeration layers are cleared at once. This comparison costs a minute and requires no change to the system, which is why it comes before any update or reinstallation. It is also the test that prevents the most common waste of effort in this area: reinstalling a driver to fix a problem that lives in the reader.
Step 3: Check the GPU-Z Version
Open GPU-Z, note its version number, and compare it with the current official version on the TechPowerUp website. If the installed version is older than the card's release year, update GPU-Z to the current version and try again.
This step costs nothing, requires no reinstallation, and carries no risk, which is why it comes before any driver work. A new card's device ID simply does not exist in an old copy's database, and no amount of driver reinstallation will put it there. Conversely, if GPU-Z is already current and still cannot recognise the card, that result is itself evidence: the version layer is cleared, and the diagnosis moves on to the driver and VBIOS layers with one candidate eliminated.
Record the version number before and after the update, in the same note started in Step 1. A version number is a fact; "I updated it" is not, and the next person to look at the diagnosis needs the number to know whether the update actually happened.
Step 4: Check the Driver State
In Device Manager, open the card's Properties, go to the Driver tab, and note the driver provider: NVIDIA, AMD, or Intel means the manufacturer's driver; Microsoft means the generic driver.
If the provider is Microsoft, install the manufacturer's driver for the exact card model, from the manufacturer's own website, and then open GPU-Z again and compare. Do not guess a driver from the generic name shown. The generic name is a clue that the manufacturer's driver is missing, not a description of the card. The driver must match the card's actual model, confirmed from the card itself or from the manufacturer's specification page.
If the manufacturer's driver is already installed, note its version number as well. A driver version is a fact that belongs in the record, alongside the GPU-Z version and the Device Manager name. If GPU-Z reads the name but all sensor readings are 0, the driver is the suspect, and a clean reinstall is the comparison test — but that test is run as a single change, with GPU-Z reopened immediately afterwards, so its result can be attributed to the reinstall and nothing else.
Step 5: Check the VBIOS and the Device ID
In GPU-Z, look at two fields: "BIOS Version" and "Device ID". Compare both with the card manufacturer's official specification page.
- If the BIOS Version field is blank or garbled, the VBIOS cannot be read. That points at the VBIOS layer, and the VBIOS diagnosis is the next guide.
- If the device ID is unrecognised, or the model name shown does not match the official specification, the possibility of a modified BIOS must be considered. A card whose BIOS has been changed to report a false model name will show that false name in GPU-Z, because GPU-Z reads what the VBIOS writes. The name on screen is not proof of the model. The device ID and the official specification page are the check.
- If the card has a dual BIOS switch, switch to the other BIOS position and read again. One side may be readable when the other is not, and the comparison itself is the evidence.
This step is about identity, not about sensors. A BIOS Version field that is blank is a different symptom from a temperature reading that is 0, and the two belong to different diagnoses. Keep them separate: this guide is about the card's identity being unreadable, and a sensor reading of 0 with the identity intact is a sensor problem.
Step 6: Isolate a Single Variable
Change only one thing at a time, and reopen GPU-Z immediately after each change. Update GPU-Z, then check. Install the driver, then check. Switch the BIOS position, then check. Never update GPU-Z, reinstall the driver, and switch the BIOS all at once.
The discipline matters because each change has a different cost and a different risk. Updating GPU-Z is free and risk-free. Installing a driver is a larger change, involving a restart and, for a clean reinstall, safe mode and a cleanup tool. Switching a BIOS position changes the card's own firmware side. Doing all three at once and seeing GPU-Z suddenly read the card proves nothing about which of the three was responsible, and the next time the problem appears, there will be no record of which change mattered. Keep a note of each single change and its result, in the same record started in Step 1.
Step 7: Cross-Test and the Stop Line
Two cross-tests complete the diagnosis:
- Test the card in a different computer. Does the other computer's GPU-Z recognise it? If it does, the problem is in the first computer — its driver, its Windows installation, or its GPU-Z copy — not in the card.
- Test a different card, known to be good, in this computer. Does this computer's GPU-Z recognise it? If it does, the problem is in the first card — its VBIOS, or its device ID — not in the computer.
The stop line is reached when any of the following is true: GPU-Z cannot read the card and Device Manager cannot recognise it either; the computer has no picture at all at start-up; there is a smell of burning; or all seven steps have been worked through, one variable at a time, and GPU-Z still cannot read the card. At that point, stop circling in the software layers. Bring the machine in with the record from Step 1: the name Device Manager shows, the GPU-Z version number, the driver version, which GPU-Z fields are blank or 0 or Unknown, and the comparison result from the second tool in Step 2. In-store inspection is charged at the site's published three-tier fees of HK$400–2,800, quoted after inspection. Inspection takes place at Room 1209, Dali Plaza, Shek Mun. The shop can be called on 2186 8010, or reached on WhatsApp at 6558 6806. There is no home-visit service.
A Quick Comparison: The Four Decision Points at a Glance
| Decision point | What it controls | How to test it | What a pass looks like |
|---|---|---|---|
| Device Manager — is the card there, and what is it called? | Whether Windows enumerates the card at all | Open Device Manager, Display adapters, record the exact name and any warning | Card listed with its correct name and no warning |
| GPU-Z version — does this copy know this card? | Whether the reader's database contains the card's device ID | Note the GPU-Z version; compare with the current official version on TechPowerUp | Current version recognises the card, or the version layer is cleared |
| Driver state — generic, manufacturer, or damaged? | Whether the driver reports the card's identity | Driver tab in Device Manager: provider is NVIDIA, AMD, Intel, or Microsoft | Manufacturer's driver installed; GPU-Z reads name and sensors |
| VBIOS and device ID — can the BIOS be read? | Whether the card's own identity can be read | GPU-Z BIOS Version and Device ID fields against the official specification | BIOS Version present and legible; device ID matches the specification |
Boundaries: Problems That Look Similar but Are Not This One
A diagnosis is only as good as its boundaries. Several nearby problems share a GPU-Z window, a blank field, or a strange name, but each belongs to a different guide, and the steps above will not fix them:
- The laptop has two graphics processors — an integrated one and a discrete one — and GPU-Z has a device selector at the top of its window. Selecting the wrong device shows the other card's information. That is a wrong selection, not a failure to read. The discrete card can also be hidden in power-saving mode, and may only appear when the laptop is plugged in and set to high performance mode. Check the selector before concluding that anything is unreadable.
- The graphics card is passed through to a virtual machine, or running without a display connected. GPU-Z may not be able to read the card in those conditions, and that is an expected limitation, not a fault. This guide covers a normal desktop computer with a display connected.
- GPU-Z reads the name correctly, but every sensor reading is 0. That is a different symptom: the identity is readable and the sensors are not. It points at a damaged driver or at the sensor layer, and it follows the sensor diagnosis, not this guide.
- GPU-Z shows a model name, but the name is wrong — a card reporting a model it is not. That is different from Unknown. Unknown means the card cannot be recognised; a wrong name means it has been recognised incorrectly. A card whose BIOS has been modified to report a false model follows the modified-BIOS diagnosis. This guide uses only the principle that a displayed name should never be trusted on its own.
- Device Manager shows the card with a yellow warning and an error code. The error code divides the diagnosis: Code 43 has its own guide, and other error codes have theirs. This guide covers the case with no error code and a purely unreadable identity.
- There is no card listed in Device Manager at all. That is an enumeration problem — Windows does not see the card — and it follows the card-not-recognised diagnosis. GPU-Z cannot read a card that Windows cannot see, and updating GPU-Z will not change that.
- The memory capacity shown does not match the specification, while the name reads correctly. That is a capacity question, not an identity question. GPU-Z reads the capacity number the BIOS writes down; it does not count memory chips directly. A capacity mismatch has its own guide, and this guide mentions only the one sentence that GPU-Z reads the BIOS value rather than measuring the hardware.
- Readings differ between two tools — GPU-Z says one temperature and another tool says a different one. That is a cross-tool difference, and it has its own guide. A difference between tools is not the same as one tool reading nothing at all.
The principle is to isolate one thing at a time. A symptom that appears only in GPU-Z, only with an old version, only under the generic driver, or only on one BIOS position is evidence — and the evidence points in different directions depending on when it appears. The seven steps are arranged so that each result narrows the field, and the note kept along the way is what turns a confusing blank field into a diagnosis someone else can continue.
How GPU-Z Reads a Card — and Why That Decides the Diagnosis
It helps to see the reading chain as a line of four stations, each handing information to the next. Windows enumerates the card and gives it a place in Device Manager. The driver talks to the card and reports what it finds. GPU-Z looks up the card's device ID in its own database. And the VBIOS supplies the identity — the model name, the memory capacity, the BIOS version — that the other stations pass along.
A break at any station changes what GPU-Z can show, and the change is characteristic. A break at enumeration leaves no card in Device Manager and nothing for GPU-Z to open. A break at the driver leaves the generic name in Device Manager and a GPU-Z window with little or nothing filled in. A break at the version leaves Device Manager correct and the driver correct, but GPU-Z unable to name a card that is newer than its database. A break at the VBIOS leaves the name present but the BIOS Version field blank, or the device ID unrecognised.
That is why the order of the steps matters, and why the order is the diagnosis itself. Device Manager first, because it tests the earliest station. A second tool next, because it tests whether the failure is in the reader or in the source. The version next, because it is free and risk-free. The driver next, because it is the largest change. The VBIOS last among the software layers, because it is the card's own identity. Each step clears one station, and the station that remains uncleared at the end is where the fault lives.
There is a second reason the frame matters: it sets the right expectation for what a fix looks like. A fix at the version station is an update to the reader. A fix at the driver station is an installation. A fix at the VBIOS station is a firmware matter. None of those is a repair to the graphics processor, and none of them requires the card to be replaced. Replacing the card to fix a reader's database, or a generic driver, is the most expensive possible answer to a question that had a free one.
Six Common Misconceptions
Misconception 1: "GPU-Z shows Unknown, so the graphics card must be broken."
No. GPU-Z is a reader, not a measurer. A blank field or an Unknown only shows that one link in the reading chain has broken, and which link it is has to be checked link by link. In many cases the cause is a driver that is not properly installed, or a copy of GPU-Z that is older than the card. The card itself can be perfectly sound while the reader cannot describe it.
Misconception 2: "The name GPU-Z shows is the card's true model."
No. GPU-Z reads the number the VBIOS writes down; it does not measure the hardware directly. If the VBIOS has been modified, GPU-Z will faithfully report the wrong name. The model name must be checked against the official specification page and the device ID. A name on screen, on its own, is not proof of the model — and a name that matches the box is not proof either, if the BIOS was changed to match the box.
Misconception 3: "Device Manager recognises the card, so GPU-Z must recognise it too."
No. Device Manager recognising the card only shows that enumeration is normal. GPU-Z still has to pass through the driver reporting layer and its own database layer. A generic driver, or a copy of GPU-Z older than the card, can leave Device Manager correct and GPU-Z unable to read the card. The two tools read different stations of the same chain, and a pass at one station does not guarantee a pass at the next.
Misconception 4: "All GPU-Z readings are 0, so the card must have no power."
No. Readings of 0 with the name still present most often point at a damaged driver, which leaves GPU-Z unable to ask the sensors anything. A card that is producing a picture on the screen already proves it has power. "Cannot read a reading" and "no power" are two different claims, and the picture on the screen settles the second one. The first one belongs to the driver and sensor layers.
Misconception 5: "Updating GPU-Z is useless — just reinstall the driver, it is faster."
No. Updating GPU-Z is the cheapest, fastest, and lowest-risk step available, and it belongs before any driver reinstallation. A driver reinstall involves safe mode, a cleanup tool, and keeping Windows Update from installing its own driver first — a much larger job. Getting the order wrong means doing a large job that may have been unnecessary, and changing a working layer on a guess. The cheap test comes first precisely because it can clear the problem in a minute.
Misconception 6: "GPU-Z cannot read the card, so replacing the card will fix it."
No. GPU-Z failing to read a card has four decision points, and replacing the card only answers one of them — the possibility that the card itself is faulty — and it is the most expensive possibility of the four. Before any replacement, the Device Manager comparison, the version comparison, and the driver comparison should all have been done. A card replaced to fix a reader's database, a generic driver, or a VBIOS that could have been read from its other side is a card replaced for nothing.
When to Stop and Bring the Machine In
GPU-Z reading failures are narrowed down by the seven steps above — often at Step 1, where Device Manager names the generic driver, or at Step 3, where an old version of GPU-Z is found, or at Step 2, where a second tool shows that only GPU-Z is failing. If all seven steps have been worked through, one variable at a time, and GPU-Z still cannot read the card — or if Device Manager cannot recognise the card either, or the computer has no picture at all, or there is a smell of burning — stop. Further guessing in the software layers risks changing settings that were never part of the problem, and makes the eventual diagnosis harder rather than easier.
Bring the machine in with the notes from Step 1 and the results of each step tried. In-store inspection takes place at Room 1209, Dali Plaza, Shek Mun. You can call 2186 8010, or reach the shop on WhatsApp at 6558 6806. Inspection is charged at the site's published three-tier fees of HK$400–2,800, quoted after inspection. There is no home-visit service; inspection is carried out in the shop.
One last point about expectations: a diagnosis that ends at the stop line is not a failed diagnosis. A note that says "Device Manager shows the correct name; GPU-Z is the current version; manufacturer's driver installed; BIOS Version field blank; Task Manager recognises the card" is a precise description of where the fault lives, and it is exactly what an in-store inspection needs to start from the right place instead of from the beginning.
FAQ
Q1: GPU-Z shows Unknown. Does that mean my graphics card is broken?
No. GPU-Z is a reader, not a measurer. An Unknown only shows that one link in the reading chain has broken — enumeration, the driver, the GPU-Z version, or the VBIOS — and which link it is has to be checked link by link. In many cases the cause is a driver that is not properly installed, or a copy of GPU-Z older than the card, and the card itself is perfectly sound.
Q2: The name GPU-Z shows must be the card's true model. Is that right?
No. GPU-Z reads the number the VBIOS writes down; it does not measure the hardware directly. If the VBIOS has been modified, GPU-Z will faithfully report the wrong name. The model name must be checked against the official specification page and the device ID. A name on screen, on its own, is not proof of the model.
Q3: Device Manager recognises the card, so GPU-Z must recognise it too. Is that true?
No. Device Manager recognising the card only shows that enumeration is normal. GPU-Z still has to pass through the driver reporting layer and its own database layer. A generic driver, or a copy of GPU-Z older than the card, can leave Device Manager correct and GPU-Z unable to read the card.
Q4: All GPU-Z readings are 0. Does that mean the card has no power?
No. Readings of 0 with the name still present most often point at a damaged driver, which leaves GPU-Z unable to ask the sensors anything. A card producing a picture on the screen already proves it has power. "Cannot read a reading" and "no power" are two different claims, and the picture settles the second one.
Q5: Updating GPU-Z is useless — I should just reinstall the driver, it is faster. Is that right?
No. Updating GPU-Z is the cheapest, fastest, and lowest-risk step available, and it belongs before any driver reinstallation. A driver reinstall involves safe mode, a cleanup tool, and keeping Windows Update from installing its own driver first. Getting the order wrong means doing a large job that may have been unnecessary.
Q6: GPU-Z cannot read the card, so replacing the card will fix it. Is that the right conclusion?
No. GPU-Z failing to read a card has four decision points, and replacing the card only answers one of them — the possibility that the card itself is faulty — and it is the most expensive possibility. Before any replacement, the Device Manager comparison, the version comparison, and the driver comparison should all have been done.
Q7: How do I tell whether it is a version problem or a driver problem?
Check the GPU-Z version first, against the current official version on the TechPowerUp website. If the installed version is older than the card's release year, update GPU-Z and try again — if it reads the card after the update, it was a version problem. If it is already the latest version and still cannot read the card, it is not a version problem: check the driver state next. If the driver provider is Microsoft's generic driver, install the manufacturer's driver; if the manufacturer's driver is already installed but the readings are all 0, that points at a damaged driver.
Q8: How should the second-tool comparison be done?
Compare with a second tool: Task Manager under Performance and then GPU, or dxdiag under its Display tab. If the second tool also cannot recognise the card, the problem is in the enumeration or driver layer, not in GPU-Z. If the second tool recognises the card and only GPU-Z cannot, the problem is locked to GPU-Z itself — its version or its database — and the other layers are cleared at once.
Q9: My laptop has two graphics cards. How do I read GPU-Z correctly?
GPU-Z has a device selector at the top of its window — before opening GPU-Z, confirm which card is selected, the integrated graphics or the discrete card. The discrete card may be hidden in power-saving mode, so plug the laptop in and switch to high-performance mode before it appears. Selecting the wrong device, or a discrete card that has not woken up, is not GPU-Z failing to read — do not follow this guide's steps for that.
Q10: When should I stop trying software fixes and bring the machine in for inspection?
Stop when GPU-Z cannot read the card and Device Manager cannot recognise it either; when the computer has no picture at all at start-up; when there is a smell of burning; or when all seven steps have been worked through, one variable at a time, and GPU-Z still cannot read the card. Bring your notes: the name Device Manager shows, the GPU-Z version number, the driver version, which GPU-Z fields are blank or 0 or Unknown, and the second tool's comparison result. In-store inspection is at Room 1209, Dali Plaza, Shek Mun, charged at the site's published three-tier fees of HK$400–2,800, quoted after inspection. You can call 2186 8010 or WhatsApp 6558 6806.