GPU Fans at Full Speed on Boot? POST Self-Test vs Stuck Full Speed — Thermal Control & Sensor Fault Diagnosis

Key Takeaways
A brief full-speed spin at power-on is usually a self-test. Only speed that stays at maximum after Windows loads needs diagnosis: compare temperature against fan speed to separate a locked setting, a faulty sensor, a broken control path, and genuine overheating.
The first few seconds: why fans spin up to maximum at power-on
Many users first notice this at the moment they press the power button: the graphics card fans immediately spin at maximum speed, loudly, for a few seconds, and then quieten down. This is usually nothing to worry about. At power-on there is a handover gap: the driver has not loaded yet, the fan controller has not received a formal speed-control signal, and the card firmware often falls back to a safe default of full speed, preferring a loud fan to no fan at all. Once the system starts and the driver takes over, the controller begins adjusting speed by temperature, and the fans settle down.
The first dividing line, therefore, is not whether full speed happened, but how long it lasted and at which stage it stopped. A brief full-speed burst that settles is in the self-test family of normal behaviour. Full speed that continues from power-on through Windows login and into an idle desktop is the pattern worth investigating, whether the cause is a setting, a sensor, or a control fault. Mixing the two together leads to two opposite errors: a normal self-test mistaken for a fault, or a real fault dismissed with "they all do that at boot."
One detail to settle first is which fans are actually noisy. Case fans and graphics card fans can both run at full speed at boot, and they sound different and come from different places. Before concluding that the graphics card is at fault, watch during boot with the side panel off, if that is safe and practical, and confirm which set of fans is spinning hardest. Never put a hand near a spinning fan while doing this.
Four causes of stuck full speed: classify before acting
Stuck full speed is not a single fault. It is an umbrella for four different causes, and acting before classifying them risks replacing the wrong part or changing the wrong setting.
The first cause is a locked setting or software conflict. A fan curve fixed at 100%, a third-party fan-control program fighting the driver control, or a leftover setting from an earlier configuration can all hold the fans at maximum. The signature is a fan that will not slow down even though temperatures are low, and that usually responds immediately when the setting is changed. Detailed fan-curve tuning belongs to a separate guide; this article uses only one diagnostic principle from it: restore the default or automatic curve and see whether full speed disappears.
The second cause is a faulty sensor reading. The card decides fan speed from temperature sensors, and if a sensor reading is distorted — persistently high, jumping erratically, or frozen at one value — the controller will run the fans at full speed on a false high-temperature signal. This is the easiest cause to misread, because on the surface it looks identical to genuine overheating. The difference lies only in whether the temperature reading matches the actual load: a computer sitting idle, running no game and no compute workload, yet reporting a persistently high temperature, points toward the sensor stage.
The third cause is a control-path fault. The speed-control signal travels through the controller, the connector, and the wiring to reach the fan. If that signal cannot get through, or the connector makes poor contact, the fan receives no slow-down instruction and stays at its safe full-speed default. The signature is a fan that does not respond to software at all: move the curve up or down and the speed does not move.
The fourth cause is genuine cooling demand. If the cooling path has failed — blocked case airflow, heavy dust on the heatsink, or poor contact between heatsink and chip — temperatures really are high, and full speed is the correct response. The signature is temperature and speed rising together, with temperature climbing sharply as soon as load is applied. Mistaking this for a sensor fault and forcing the speed down would only make temperatures higher, so this cause must be separated carefully from the second one.
| Cause | Typical signature | First response |
|---|---|---|
| Locked setting / software conflict | Temperature low but speed at maximum; changing the setting has an immediate effect | Restore the default curve; close third-party fan-control software |
| Faulty sensor reading | Temperature reading does not match load; persistently high or frozen | Record idle and loaded readings and compare the change |
| Control-path fault | No response to any software change | Inspect the connector and wiring; do not keep adjusting software |
| Genuine cooling demand | Temperature and speed high together; temperature rises sharply under load | Check airflow and heatsink contact; do not force the speed down |
Diagnosis: use the timeline and the temperature-speed pairing
Step one is to record the timeline. How many seconds does full speed last after power-on? What happens at the login screen? What happens after ten idle minutes on the Windows desktop? Full speed confined to the early boot stage, followed by a settle, belongs to the self-test pattern and needs no further action. Full speed that persists into Windows is the trigger for the next steps.
Step two is to read temperature and speed together. Open a monitoring tool and record idle temperature and speed, then record the change under a light load. An introductory guide to reading monitoring tools covers the full set of readings; this article needs only two numbers and their pairing. Low temperature with full speed points to a setting, sensor, or control problem. High temperature with full speed points to genuine overheating or a sensor reporting high, and step three separates those two.
| Idle state | Points toward | Next step |
|---|---|---|
| Low temperature, full speed | Locked setting, faulty sensor, or control fault | Restore default settings and watch for any speed response |
| High temperature, full speed | Genuine overheating or a sensor reporting high | Apply a light load and check whether the temperature rise is plausible |
| Low temperature, speed settles | Normal boot self-test | No action needed; watch for future change |
| Temperature reading frozen | Sensor stage is suspect | Record readings across several sessions to hand over at inspection |
Step three is to rule out software. Restore the default fan curve, close every third-party fan-control program, and if needed reinstall the driver cleanly, then observe once more. Change only one variable at a time: if the curve, the driver, and the dust are all changed together and the problem goes away, nobody knows which change mattered, and a recurrence will have to be diagnosed from scratch.
Step four is to separate one fan from all fans. If only one fan runs at full speed while the others behave normally, the fault is more likely in that fan or its wiring. If all fans run at full speed together, the fault is more likely in the shared control, sensor, or setting. This distinction matters because the inspection direction is different for the two patterns.
Step five is listening and looking, without opening the card. If full speed comes with grinding, knocking, or obvious vibration, a mechanical problem becomes more likely, and sound classification belongs to the fan-noise guide rather than to temperature-control logic. Also check for foreign objects touching the blades and for case airflow blocked by cables or clutter. None of these checks requires opening the graphics card heatsink.
Boundary cases: when full speed is not the real problem
Full speed combined with no display, or with a boot that never completes, may be only a side effect: the card has not finished initialising, so the controller stays at its safe default. In that case the boot failure itself must be addressed first — power connectors, slot seating, and output connections — rather than chasing fan quietness. A quiet fan on a machine that cannot boot is not a fix.
On cards with a fan-stop design, a full-speed self-test followed by fans stopping at idle is normal behaviour, and the stop mechanism is covered by the fan-stop guide. Do not mistake "spin up, then stop" for a fan that works intermittently, and do not treat the self-test burst as an early warning simply because the fans later stop. Both are two sides of the same logic: spin when needed, stop when not.
Water-cooled cards need separate treatment. Pump speed and radiator-fan control follow different logic from air-cooled cards, and a pump problem, restricted coolant flow, or radiator-fan setting can produce loud noise or high temperatures that do not map onto the air-cooled diagnosis in this article. This guide covers air-cooled cards; water-cooled owners should check the pump and radiator as a separate system.
Some cards have a dual-BIOS switch, and the two positions can use quite different fan strategies. A change in full-speed behaviour after switching positions does not mean one position is broken; it may simply be a different strategy. Switching BIOS positions and flashing firmware both carry risk, so without confidence it is better to record the behaviour of both positions and hand that record over at inspection.
Common misconceptions, taken apart
Misconception one: full speed means the card is about to fail. A brief self-test burst is designed behaviour, not a warning. What deserves attention is sustained full speed combined with abnormal temperatures, or sustained full speed that software cannot influence at all.
Misconception two: cleaning dust always fixes full speed. Dust affects cooling efficiency and can push temperatures up, but it does not directly lock a controller at 100%. When temperatures are low and speed is still at maximum, dust removal will not change the speed, and the setting and sensor stages should be checked first.
Misconception three: lowering the curve counts as a repair. If the sensor or control path is faulty, forcing the curve down only hides the signal; the temperature information itself is already wrong, and suppressing speed when genuine overheating occurs is dangerous.
Misconception four: replacing the fan always fixes it. If the fan itself is healthy and the fault sits in the sensor, control, or setting, a new fan will receive the same wrong instruction and run at full speed too. Replacement should wait until the fault has been shown to sit in the fan itself. Fan replacement, including model matching and removal risks, belongs to its own guide.
Misconception five: a zero speed reading means the fan has stopped. Speed sensing and fan power are separate signals: a fan can be spinning while its speed-sense line is faulty and the reading shows zero. Visual confirmation must never be skipped, and a single number alone should not decide the diagnosis.
When to stop and send the card for inspection
Stop dismantling or flashing firmware if any of these appear: speed shows no response at all after settings are restored to defaults; the temperature reading stays frozen or bears no relation to load; one fan runs at full speed and its connector already shows visible damage; or full speed at boot comes with no display or a failed boot. Bring three records to inspection: the full-speed timeline, the idle and loaded temperature and speed readings, and a list of which settings were changed and with what result. Those records let inspection aim at the right stage immediately.
For inspection arrangements, the shop is at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806. Repair charges follow the published three-tier range of HK$400–2,800, with the actual amount confirmed by a written quotation after inspection; no chargeable repair is carried out before that quotation.
Frequently asked questions
Q: My fans spin at full speed for a few seconds at boot, then go quiet. Is that normal?
A: In most cases, yes. Before the driver takes over, the controller falls back to a safe full-speed default, and it settles once the system and driver are running. If idle speed and temperature are normal afterwards, no action is needed.
Q: Windows has loaded and the fans are still at full speed, but temperatures are low. What is most likely?
A: Most likely a locked setting or a sensor or control problem. Restore the default fan curve and close third-party fan-control software first: a response points to settings, while no response at all points toward the sensor or control path.
Q: Temperatures are high and the fans are at full speed. Can I just lower the speed?
A: That is not advisable. When temperatures are genuinely high, full speed is the correct response, and forcing it down will push temperatures higher. Find why temperatures are high — airflow, heatsink contact, or load source — and the speed will fall on its own once temperatures do.
Q: Only one fan runs at full speed while the others are normal. What does that mean?
A: It points more toward that individual fan or its wiring than toward a shared setting. Check for objects touching that fan and for a loose connector before concluding that the whole card control has failed.
Q: The speed reading shows zero, but the fan is clearly spinning. Why?
A: Speed sensing is a separate signal from fan power. A faulty sense line can report zero while the fan keeps spinning. Confirm the fan state visually, and report the difference between the reading and what was seen when handing the card over.
Q: I cleaned the dust and the fans are still at full speed. Did the cleaning fail?
A: Not necessarily. Cleaning addresses cooling efficiency; if the original cause was a setting, sensor, or control fault, cleaning was never going to change the speed. That result is itself diagnostic information pointing away from cooling efficiency and toward settings and sensing.