GPU Fan Spinning but RPM Reads Zero? Tach-Signal vs Fan-Controller Fault Diagnosis

| HK GPU Fix

GPU Fan Spinning but RPM Reads Zero? Tach-Signal vs Fan-Controller Fault Diagnosis

Key Takeaways

If a graphics card fan is visibly spinning yet monitoring software reports 0 RPM, the fan motor is not necessarily faulty: the speed-reporting tach signal may be missing while the drive signal still works. This guide explains how to match each fan to its reading, cross-check software, test the fan curve response, and decide when to stop and seek inspection.

What This Article Covers, and What It Does Not

This article is for one specific mismatch: you can see a graphics card fan spinning with your own eyes, but monitoring software reports that fan at 0 RPM, or shows a reading that does not move. The graphics card is producing a picture and is running. The problem is not that the fan has stopped. The problem is that what the software reports and what the eye sees do not agree.

That distinction defines the whole article. A fan that has genuinely stopped is a different problem, covered by other guides on this site. A fan that is spinning at full speed uncontrollably is also a different problem. A fan that stops at idle by design, under a zero-RPM idle mode, is also different: in that case the fan is genuinely not spinning, and a 0 RPM reading is correct. This article applies only when the fan is genuinely spinning and the reading says zero. Getting that entry condition right matters, because every step that follows assumes it. If the fan is not actually spinning, the steps below will mislead.

The article also does not teach fan replacement, fan curve design, or temperature-sensor diagnosis. Those are separate subjects. Fan curve controls are used here for one diagnostic purpose only: to test whether the drive side still responds. Temperature readings are a separate sensing channel, and a temperature fault is diagnosed by different methods. Keeping those boundaries clear prevents a speed-reporting fault from being confused with a cooling fault, a temperature-sensor fault, or a fan-motor fault, all of which look different once the right test is applied.

Two Separate Paths: Driving a Fan and Reporting Its Speed

A graphics card fan system has two independent paths, and confusing them is the root of most misunderstanding about a 0 RPM reading.

The first path is the drive path. A control signal, commonly a pulse-width modulation signal, tells the fan how fast to spin. The fan controller varies that signal according to temperature, load, and the configured fan curve. When the drive path works, the fan speeds up and slows down as commanded. The drive path is about causing motion.

The second path is the reporting path. As the fan spins, it sends back a tachometer signal — a series of pulses, typically two pulses per revolution — that reports its actual speed to the controller and, through it, to monitoring software. The software counts those pulses over time and calculates a revolutions-per-minute figure. The reporting path is about measuring motion that has already happened.

Those two paths can fail independently. If the drive path fails, the fan does not spin as commanded, or does not spin at all. If the reporting path fails, the fan spins exactly as commanded, but the controller and the software receive no pulses, so they calculate 0 RPM. The fan is not broken in the sense that matters for airflow. The reporting is broken. That is the central mechanism of this article, and it explains why "the fan is spinning, so the reading must be wrong" and "the reading is zero, so the fan must be dead" can both be wrong in different ways. The fan can be alive and the reading can be genuinely zero, because the reading measures pulses, not motion directly.

Understanding that separation also explains why a 0 RPM reading is not purely a display problem. Automatic temperature control depends on feedback. When the controller receives no speed feedback, it is adjusting blind: it can still command a speed, but it cannot confirm what speed was achieved. A fan curve setting may therefore have no verifiable effect, even though the fan itself still moves air. Temperature protection itself, however, still relies on temperature sensing, which is a different channel. A 0 RPM reading does not mean temperature protection has disappeared, and it should not be described that way. It means one feedback channel is missing, and that missing feedback matters for control even when it does not mean the fan has stopped.

Where the Reporting Path Can Break

The reporting path can break at several points, and the diagnostic sequence is ordered roughly from the cheapest explanation to the more involved ones. Listing them here, before the steps, helps the reader understand why each step exists.

First, the software may be reading the wrong fan. On a multi-fan card, monitoring software shows one reading per fan position. If the reader looks at the wrong position, a spinning fan can appear to read zero while a different, stopped fan is the one actually showing zero. This is a reading error, not a hardware fault at all, and it is the first thing to exclude.

Second, some card designs only report one fan. On such a design, the main fan has a speed signal, but additional fans may have no independent reading. Their reading may show zero permanently, or may not be shown at all, even though all fans are spinning normally. That is normal for that design, not a fault, but it can only be recognised by matching each physical fan to its reading, as described in Step 1.

Third, the fan connector may not be fully seated, or a pin may be bent. The drive and reporting signals travel through the same connector, but a partially seated connector can in principle affect one signal while the other still makes contact. A loose connector is therefore a real candidate, but it is also a candidate that involves opening the card or its shroud, which carries warranty and damage considerations. That is why connector checking comes later in the sequence, with explicit conditions, rather than first.

Fourth, the tach signal wire itself may be broken. A break in that single wire leaves the drive path intact — the fan still receives its drive signal and spins — while the reporting path is cut entirely. The software then sees no pulses and reports zero, regardless of how fast the fan is actually turning.

Fifth, the fan connector or the controller reading side may be faulty. In that case the fan may be sending pulses, but the controller cannot read them, or the controller cannot pass them on. The symptom is the same — spinning fan, zero reading — but the fault is on the reading side rather than on the fan side.

Sixth, a replacement fan may have no tach signal wire at all. Some replacement fans are built without a speed-reporting wire, or with a different wiring arrangement. Such a fan will spin when driven, but will never report a speed, because it has no reporting path to report through. This is worth knowing before replacing a fan in the hope of fixing a zero reading: if the replacement fan also lacks a reporting wire, or if the fault is on the controller side, the reading will remain zero after the replacement, and the replacement will have proved nothing.

None of these possibilities can be distinguished by looking at the zero reading alone. They are distinguished by the sequence that follows.

The Diagnostic Sequence, in Order

Work through the steps in order, changing one thing at a time and recording the result. The sequence is designed so that each step excludes one explanation before the next, more involved step is attempted.

Step 1: Match Each Fan to Its Reading, One by One

Do not start with a general statement such as "the fan is spinning." Identify which physical fan is spinning, and which software reading corresponds to it. On a multi-fan card, look at each fan in turn: the left fan, the middle fan, and the right fan, or however many the card has. For each one, note whether it is spinning, and note what its corresponding software reading shows.

This step has two possible outcomes, and they lead in different directions. If the fan that is spinning is the same fan whose reading shows zero, then this article applies, and the diagnosis continues to Step 2. If the spinning fan has a normal reading, and the zero reading belongs to a different fan that is genuinely not spinning, then this article does not apply. The problem is a stopped fan, not a reporting fault, and the stopped-fan guides are the relevant ones.

On some designs, there may be no separate reading for every fan. If a particular fan has no reading at all, or if only one reading exists for a multi-fan card, note that carefully. A fan with no independent reading is not the same as a fan with a zero reading, and the conclusion is different: the design may simply not report that fan. That conclusion should only be reached after the matching exercise has been done fan by fan, not assumed in advance.

This step sounds elementary, and that is exactly why it is first. Many zero-RPM reports dissolve at this step, because the reader was looking at the wrong fan position, or because the card never reported that fan in the first place. Skipping it guarantees confusion later, when a curve test or a connector check is applied to the wrong fan.

Step 2: Cross-Check With a Second Monitoring Program

Open a second monitoring program — for example, if the first reading came from GPU-Z, check HWiNFO or the manufacturer's own software, or the reverse — and look at the same fan position. Compare the two readings for the same fan.

If both programs report zero for the same fan, a single-program reading error is excluded. The fault is not a quirk of one program's sensor selection. If one program shows a speed and the other shows zero for the same fan, the problem is at the software or sensor-selection layer: one program is reading the wrong sensor, or is failing to read a sensor that the other program reads correctly. In that case, the diagnosis points toward monitoring-software differences, a subject covered in a separate guide, and the hardware steps that follow do not apply in the same way.

Note the condition carefully: the comparison must be for the same fan position. Comparing one program's first-fan reading with another program's second-fan reading proves nothing, because multi-fan cards can report fans in different orders. Match the fan first, as in Step 1, then compare programs for that fan.

Step 3: The Fan Curve Response Test

This is the decisive test that separates a drive fault from a pure reporting fault. In the manufacturer's software, set the fan in question to a fixed, higher speed manually. Then observe two things independently: what the fan physically does, and what the reading shows.

Listen to the fan and watch its blades. If the fan audibly and visibly speeds up, the drive path is working: the controller commanded a higher speed, and the fan obeyed. If the reading nevertheless remains at zero while the fan is clearly spinning faster, the diagnosis is confirmed as a reporting fault. The drive works, the reporting does not. That combination — responsive drive, zero reporting — is the signature of this article's fault.

If, on the other hand, the fan does not speed up at all when commanded to a fixed higher speed, then the drive side is also involved. The fault is not a pure reporting fault. The drive layer needs attention, and the direction of diagnosis changes toward fan-drive and control faults covered in other guides. Do not force the conclusion back to reporting simply because the reading is zero. The curve test is designed to catch exactly that case, and its result should be respected.

Two cautions apply to this test. First, it is a diagnostic test, not a fix. Setting a higher fixed speed does not repair a broken reporting wire, and it should not be left as a permanent workaround without understanding the consequences. In particular, a fixed speed that is left in place may mask a later change in cooling need, because automatic control is no longer varying the speed with temperature. Second, perform the test at a modest fixed speed, not at maximum, and return the setting to automatic afterwards unless there is a specific reason to do otherwise. The test needs only enough change to see and hear a response.

Step 4: Fan Connector Check, With Conditions

If the curve test confirms a pure reporting fault, the next candidate is the fan connector. This step carries conditions that must be respected. It should only be attempted if the card is out of warranty, or if the reader is confident about opening the shroud, and only if no warranty sticker would be disturbed. If the card is still under warranty, if the reader is not confident about the procedure, or if a warranty sticker is visible, do not open the card. Seek inspection directly, with the results of Steps 1 to 3 recorded.

Where the conditions are met, disconnect power, and check whether the fan connector is fully seated, and whether any pin is bent or damaged. Reseat the connector firmly, reassemble, and check the reading again. If the reading returns, the fault was at the connector contact layer. If the reading remains zero after a firmly seated connector, the fault is deeper: a broken reporting wire, a controller reading fault, or a replacement fan without a reporting wire, among the possibilities listed earlier. Those cannot be separated at home without specialist work, and the diagnosis moves to Step 5.

Note what this step deliberately does not include. It does not include cutting wires, jumpering connections, modifying the fan connector, or rewiring a fan. Those are repair-level actions, not home diagnostic steps, and they carry a risk of damaging the card or creating a new fault. They are excluded from this article on purpose.

Step 5: The Inspection Boundary

If the connector is firmly seated and the reading remains zero, or if the curve test showed a drive problem as well, or if temperatures climb steadily under light load, stop home testing. Further steps require specialist equipment and carry risk. In particular, do not run a heavy load stress test to "see whether it overheats" while a reporting fault is suspected, because automatic temperature control may already be operating without speed feedback. A stress test under those conditions does not provide useful diagnostic information that outweighs the risk of overheating.

Record the results of Steps 1 to 3 for inspection: which fan was spinning, which reading showed zero, what the second program showed, and what the curve test showed. That record allows inspection to start from the confirmed reporting fault, rather than repeating the whole sequence from the beginning.

Why a Zero Reading Is Not Just a Display Quirk

It is tempting to dismiss a zero reading when the fan is visibly spinning. The reasoning runs: the fan moves air, the card is not overheating right now, so the number on the screen does not matter. That reasoning misses the control implication.

Automatic fan control is a feedback loop. The controller commands a speed, reads back the actual speed, and adjusts. Without the read-back, the loop is open. The controller can still command, but it cannot verify. A fan curve setting that depends on feedback may therefore have no verifiable effect, even if the fan still spins. The system is cooling, but it is cooling without the confirmation it was designed to have.

That does not mean temperature protection has vanished. Temperature protection relies on temperature sensing, which is a separate channel from speed reporting. A card can still respond to temperature even when it cannot report fan speed. The correct statement is narrower and more accurate: speed feedback is missing, so speed control is operating blind, while temperature sensing remains a separate channel. Both halves of that statement matter. Overstating the problem as "no protection" causes unnecessary alarm. Understating it as "just a display number" hides a real control degradation.

There is also a practical monitoring consequence. If speed reporting is broken, later changes — a fan that begins to slow, a bearing that begins to drag, a drive fault that develops — will not show up as a changing speed reading, because the reading is already stuck at zero. The monitoring that would normally catch a developing fault is already blind. That is a further reason to treat a confirmed reporting fault as a matter for inspection, even when temperatures currently look acceptable.

Comparing Related Symptoms

What is observedWhat the reading showsWhat it most likely means
Fan spinning, reading zero0 RPM for that same fanReporting-path fault, as in this article
Fan stopped, reading zero0 RPM for that fanGenuine stopped fan; a different diagnostic path
Fan stopped at idle, reading zero0 RPM, fan starts under loadNormal idle stop design, if the fan starts when load rises
Fan spinning uncontrollably fastVery high or erratic readingSpeed control fault, not a reporting fault alone
Fan spinning, reading normal for one fan but zero for anotherMixed readings across fansCheck fan-by-fan matching first; the design may report only the main fan
Temperature reading abnormal, fan reading normalTemperature stuck, jumping, or implausibleTemperature-sensor channel fault, a separate subject from speed reporting
Two programs disagree on the same fanOne shows a speed, one shows zeroSoftware or sensor-selection difference for that fan

The table is a summary, not a substitute for the sequence. In particular, the first row is only reached after Step 1 has confirmed that the spinning fan and the zero reading belong to the same fan. Without that matching, the table can mislead, because a mixed-fan situation can mimic a reporting fault when it is in fact a stopped-fan situation.

Boundaries and Stop Rules

These rules mark where home diagnosis stops.

First, if the speed reading is zero and temperatures climb steadily under light load, or if the fan's actual speed does not keep up with load, do not reassure yourself with "the fan is still spinning." Reduce load and seek inspection. A spinning fan that cannot keep up, or a system whose temperature climbs despite a spinning fan, is not a stable situation, and the reporting fault may be accompanied by a cooling problem that the missing feedback is hiding.

Second, any burning smell, scorch mark, or blackened connector means immediate stop and inspection. Do not power the system on again to re-check the reading. The physical sign is itself the evidence.

Third, do not run a heavy load stress test while a reporting fault is suspected, in order to see whether the card overheats. Automatic control may already be blind, and a stress test under those conditions risks overheating for no diagnostic gain. The curve response test in Step 3, performed at a modest fixed speed, is the appropriate test. A stress test is not a substitute for it.

Fourth, do not cut, jumper, or rewire fan connections at home. Those actions belong to repair work, not to diagnosis, and they can create new faults or invalidate a warranty. If a wiring fault is suspected after the connector check, that suspicion goes to inspection.

Fifth, respect the warranty condition in Step 4 without exception. If the card is under warranty, if opening it would disturb a warranty sticker, or if the reader is not confident, the connector check is skipped and inspection is sought with the results of Steps 1 to 3. A diagnostic step that voids a warranty is not a saving; it is a cost.

Common Misconceptions

"A zero reading means the fan is dead and must be replaced"

If the fan is visibly spinning, the motor is not dead. What is missing is the speed report. Replacing the fan may not fix the reading at all, if the fault is in the reporting wire, the connector, or the controller side. Replacement is a conclusion that follows diagnosis, not a first step that substitutes for it.

"The software is wrong, so it can be ignored"

A single-program reading error is a real possibility, and Step 2 exists to exclude it. But if two programs agree on zero for the same fan, the software explanation is excluded, and ignoring the reading means ignoring a confirmed reporting fault with control implications. The distinction between "one program is wrong" and "the reporting path is broken" is exactly what the cross-check step establishes, and it should not be skipped.

"A multi-fan card should show a reading for every fan"

Not necessarily. Some designs report only the main fan, and other fans have no independent reading. A zero or absent reading for a secondary fan may be normal for that design. That is why Step 1 matches each fan to its reading before any conclusion is drawn. Assuming that every fan must have a reading leads to false fault reports on cards that are working as designed.

"Replacing the fan will definitely fix a zero reading"

Not necessarily, for two reasons. First, the fault may be on the controller side, in which case any fan will still read zero. Second, a replacement fan may itself have no speed-reporting wire, in which case it will spin but will never report a speed. Before replacing a fan for a reporting fault, the replacement fan's reporting arrangement and the controller side both need to be considered, and the curve test result should be known. Otherwise the replacement proves nothing and may be blamed for a fault it did not cause.

"Setting a higher fan curve fixes the zero reading"

A fan curve controls the drive path. It commands a speed. It does not repair the reporting path. The curve test in Step 3 uses a fixed speed as a diagnostic tool: it shows whether the drive responds while the reporting stays at zero. That diagnostic use is valuable. But leaving a higher fixed speed in place as a supposed fix confuses a test with a repair, and it removes automatic variation with temperature. The reading will still be zero, because the reporting wire or controller fault has not been addressed.

"Zero RPM means there is no temperature protection"

No. Temperature protection relies on temperature sensing, which is a separate channel from fan speed reporting. A zero speed reading means speed feedback is missing. It does not mean temperature sensing has failed. Conflating the two channels leads to wrong conclusions in both directions: it can cause alarm when the card is in fact still sensing temperature, and it can hide a genuine temperature-sensor fault that happens to coincide with a speed-reporting fault. Keep the channels separate.

"There must be a single normal RPM value to compare against"

There is no universal normal speed. Fan speed varies with the card design, the temperature, the load, and the configured curve. A speed that is normal for one card under one load may be unusual for another. That is why this article does not state a normal RPM figure, and why the diagnostic tests compare a fan with itself — its physical motion against its own reading, and its response to a commanded change — rather than comparing its speed with a generic number.

What to Bring to Inspection

If the sequence ends at the inspection boundary, bring a clear record: which fan was spinning and which reading showed zero, confirmed fan by fan; what the second monitoring program showed for the same fan; what the curve response test showed, in terms of both physical response and reading; whether the connector was checked and what was found, or why it was not checked; and whether temperatures climb under light load. Also note any burning smell or connector discolouration, and do not power the card on again if either is present.

Inspection charges follow the site's published three-tier range of HK$400–2,800, with the actual quote given after inspection. No other price is stated in this article, because the correct charge depends on what inspection finds.

Frequently Asked Questions

The fan is spinning but the software says 0 RPM. Is the fan broken?

Not necessarily, and in the sense meant by this article, usually not. If the fan is visibly spinning, the motor is turning. What is missing is the speed report. The drive path and the reporting path are separate, and a fault in the reporting path leaves the fan spinning while the software sees no pulses and calculates zero. The diagnostic sequence establishes whether that is the case, rather than assuming it from the reading alone.

How do I know which fan the zero reading refers to?

Match fans one by one. Look at each physical fan and note whether it is spinning, then look at the software reading for each fan position in turn. Only when the spinning fan and the zero reading have been matched to the same fan does this article apply. If the spinning fan has a normal reading and a different fan is stopped, the problem is a stopped fan, not a reporting fault.

Two programs give different readings for the same fan. Which one is right?

Neither can be assumed right without further checking. First, confirm that both programs are reporting the same fan position, because fan order can differ between programs. If they genuinely disagree on the same fan, the problem is at the software or sensor-selection layer, and the hardware steps in this article do not apply in the same way. That software-difference subject is covered in a separate guide.

I set a fixed higher fan speed and the fan sped up, but the reading stayed at zero. What does that mean?

That is the signature of a pure reporting fault. The drive path works, because the fan obeyed the command and sped up. The reporting path does not, because the reading stayed at zero despite faster spinning. That combination confirms the fault is in reporting, not in driving, and it moves the diagnosis toward the connector and deeper reporting-path causes.

I set a fixed higher fan speed and the fan did not speed up at all. What does that mean?

That means the drive side is also involved, and the fault is not a pure reporting fault. The fan did not obey the command, so the drive layer needs attention in addition to, or instead of, the reporting layer. The direction of diagnosis changes toward fan-drive faults, and the pure-reporting steps that follow should not be forced onto a drive problem.

My card has three fans but only one speed reading. Is that a fault?

Not necessarily. Some designs report only the main fan, and other fans have no independent reading. That arrangement is normal for those designs. The way to tell is the fan-by-fan matching in Step 1: if only one reading exists and it corresponds to the main fan, while the other fans spin normally, the absence of separate readings is a design characteristic, not a fault.

Will replacing the fan fix a zero reading?

It may, but only if the fault is in the fan's own reporting wire and the replacement fan has a working reporting wire. If the fault is in the connector, the controller side, or the reporting wire path outside the fan, a replacement fan will still read zero. If the replacement fan itself has no reporting wire, it will also read zero. Replacement should follow the diagnostic sequence, not precede it.

Can I just leave the fan on a fixed high speed and ignore the zero reading?

That is not recommended as a solution. A fixed speed is a diagnostic tool in this article, not a repair. Leaving it in place removes automatic variation with temperature, and it does not restore the missing speed feedback. The control system remains blind to actual speed, and later changes in fan behaviour will not show up in the reading, because the reading is already stuck at zero.

Is it safe to keep using the card while the reading is zero, if temperatures look fine?

Temperatures looking fine at the moment does not settle the question. Speed feedback is missing, so automatic control is operating without confirmation, and monitoring is blind to later fan changes. If temperatures climb under light load, or if the fan does not keep up with load, reduce load and seek inspection promptly. Even when temperatures are currently stable, a confirmed reporting fault warrants inspection to establish which part of the reporting path has failed.

Should I open the card to check the fan connector if it is still under warranty?

No. If the card is under warranty, or if opening it would disturb a warranty sticker, or if you are not confident about the procedure, do not open it. Seek inspection with the results of the matching, software cross-check, and curve response steps recorded. A connector check that voids a warranty is not worth the diagnostic information it provides, because inspection can perform the same check without that cost.

Summary

A spinning fan with a zero reading is a mismatch between motion and measurement, not automatically a dead fan. The drive path tells the fan to spin. The reporting path tells the controller how fast it is spinning. When the drive works and the reporting does not, the fan spins and the software reports zero. Matching each fan to its reading, cross-checking software, and testing the curve response separates a reading error from a genuine reporting fault, and separates a reporting fault from a drive fault. Connector checking follows only under the right warranty conditions, and anything beyond a firmly seated connector, or any sign of climbing temperatures or physical damage, is the point to stop and seek inspection.

For inspection, HK GPU Fix is at Room 1209, Dali Plaza, Shek Mun. The telephone number is 2186 8010 and WhatsApp is 6558 6806. No home-visit service is offered. Inspection charges follow the site's published three-tier range of HK$400–2,800, with the actual quote given after inspection.

Related articles

Related services

If you need Cooling repair, see these service pages: