You drew the fan curve you wanted. Twenty per cent at idle, quiet enough to forget about. You saved, rebooted, and the fan is sitting at forty — the slider goes lower and the fan does not.
The board is not ignoring you. It is controlling the fan by a method that physically cannot go any slower, and the forty per cent is a floor it is holding deliberately so that your fan does not stop and stay stopped.
Two completely different ways to slow a fan down
Lowering the voltage
The old method, and the only one a three-wire fan can accept. The board supplies less than 12 V — 7 V, 5 V — and the fan runs slower because it has less to work with.
The catch is that everything in the fan gets less. A computer fan is brushless, which means it has electronics inside it: a Hall sensor to find the rotor and a driver to energise the coils in the right order. Those electronics need a supply to function, and the motor needs enough torque to overcome the stiction in its own bearing. Starve the voltage and you eventually reach a point where the rotor judders and stops, or spins but will not start from rest.
Now put a number on the floor. Forty per cent of 12 V is 4.8 volts. Typical 12 V fans specify a start voltage somewhere between 4.5 and 7 V. The board's floor is sitting exactly on the edge of the range where a large fraction of fans will refuse to spin up.
Pulsing the full voltage
The four-wire method, and the reason the fourth wire exists. The supply stays at 12 V and a control line switches the motor on and off very quickly — the specification puts the base frequency at 25 kHz, with 21 to 28 kHz acceptable, deliberately above hearing.
During every on-pulse the fan's electronics see the full 12 V. The sensor works properly, the driver has its supply, and the coils get full-strength kicks; the rotor's inertia smooths the pulses into steady rotation. The duty cycle sets the speed without ever telling the motor to operate on a starved supply.
That is the whole difference. Voltage control gives the fan less of everything. Pulse control gives it everything, less often.
Why the difference shows up as a floor
There is an asymmetry in fan motors that explains why the floor is set so conservatively: the voltage needed to start a fan from rest is higher than the voltage needed to keep it turning. Breaking stiction takes more torque than sustaining rotation against it.
So a board running in voltage mode has to pick a floor that works for the harder case. If it allowed you down to a level that keeps a spinning fan alive, then any event that stopped that fan — a curve that briefly dropped, a resume from sleep, a finger — would leave it stationary with a voltage too low to restart it, humming quietly and cooling nothing. Boards set the floor high enough to avoid that, and because they cannot know which fan you plugged in, they set it for the worst fan you might have.
The pulse method has a way out that voltage control simply does not: the specification explicitly permits a short burst of higher duty to get the fan moving — no more than about two seconds — after which it can settle to a duty far below anything that would have started it. Fan vendors are also required to state a minimum speed, and that minimum must be no more than 30 per cent of the fan's maximum. The low end is designed in.
Finding out which mode you are in
Four things put a header into voltage mode, and only one of them is obvious.
- The fan has three wires. There is no control line, so voltage control is the only option available.
- The header is set to DC in the firmware even though the fan has four. This is the common case and the easy fix — most boards expose a per-header mode with DC, PWM and Auto.
- Auto guessed wrong. Auto-detection probes the fan at boot and is not infallible, particularly through splitters and hubs.
- The header is a specific type — some boards wire certain headers, often pump headers, to full voltage or to a fixed mode regardless of the setting.
And one honest caveat before you go changing settings: not every floor is a mode problem. Many boards have a separate minimum-duty setting that clamps the curve independently, and some apply a step-up delay that makes a low setting look like it is being ignored when it is only being approached slowly.
What pulse control does not fix
It would be a poor article that sold you the four-wire method as free.
Some motors are audible at very low duty in a way they are not at low voltage — a faint ticking or clicking as the coils are pulsed, most noticeable in a quiet room at night, which is exactly when you were trying to make the machine silent. It is motor-dependent rather than universal, and it is the one argument in favour of voltage control that holds up.
Zero-speed modes, where fans stop entirely under a threshold, need pulse control and a fan rated for it, and they bring their own annoyance: the fan restarting audibly every time the load crosses the threshold. A slow-but-never-stopping curve is usually the quieter experience.
And a three-wire fan is not a worse fan. Plenty of excellent fans are three-wire. The floor is a property of the control method, not a verdict on the part.
The speed reading lies in two directions
The remaining wire is the tachometer, and the number it produces causes more confusion than the control method does.
The fan pulses that line as it turns, and the board counts pulses and converts them to a speed — assuming two pulses per revolution, which is the convention nearly all fans follow. Nearly. A fan that produces a different number reports a speed that is exactly double or exactly half the truth, and the reading is stable and plausible-looking while being wrong by a factor of two.
The second and more common confusion is splitters. A splitter passes the control signal to every fan, because that line is an output from the board. It cannot pass every tachometer back, because that line is an input and there is only one of it — so a well-made splitter connects the tachometer of exactly one fan and leaves the others disconnected. Three fans spinning, one number, and no indication which of them it belongs to. If a fan on that splitter dies, nothing reports it unless it happens to be the one wired through.
That matters because of what boards do with a zero reading. Many will warn, and some will refuse to boot or shut down on a fan-stop condition — which is a sensible protection and an unhelpful one if the fan is fine and the reading is missing. It is also why a fan configured below its own minimum speed can produce a fan-error alarm on a perfectly healthy machine: it has not failed, it is turning too slowly to be counted reliably.
What to change
- Set the header explicitly to PWM if the fan has four wires. Do not leave it on Auto — Auto is a guess, and this is a setting you know the answer to.
- Check whether a separate minimum-duty setting is clamping the curve before concluding anything about the mode.
- Group fans on a powered PWM hub rather than daisy-chaining several off one header. One control signal reaches all of them, only one tachometer is read back, and the header is not asked to supply current for four motors.
- If the fan is three-wire and you want it slower, either fit a low-noise adapter — a resistor in line, which lowers the ceiling rather than the floor — or replace it with a four-wire fan whose stated minimum speed is what you actually want.
- Read the fan's own minimum-speed figure before buying. It is published, it varies enormously between models, and it is the number that decides how quiet your idle can be.
Two things we would not buy
A fan controller bought to solve this. In the overwhelming majority of cases the fix is one firmware setting on the header the fan is already plugged into, and a controller adds a box, a cable run and another thing to fail.
And three-wire fans for a build whose whole point is a silent idle. They are fine parts, they simply cannot be asked to do the thing you are buying the build for, and the floor you will hit is not adjustable from software.
How this was put together
The control behaviour comes from the four-wire fan specification itself, which sets the 25 kHz base frequency and its 21–28 kHz tolerance band, requires the 12 V supply to stay at 12 V regardless of commanded speed, obliges the fan vendor to state a minimum speed no greater than 30 per cent of maximum, and permits the brief high-duty pulse used to start a stationary fan. The start-voltage ranges are the fan manufacturers' own published figures, which they state per model. The mode settings and their per-header exceptions are documented by the board vendors, and the low-duty noise behaviour is the sort of thing measured in acoustic testing by Gamers Nexus and Igor's Lab rather than claimed on a box.
The derived figure is ours: forty per cent of 12 V is 4.8 volts, which sits inside the 4.5-to-7-volt band where fans specify their start voltage — and, combined with the fact that starting a fan takes more voltage than keeping it turning, that is the whole explanation for why a board that dims instead of pulsing has to stop where it stops.
