I connected the rear fans from a Power Mac G5 (single processor, 1.8 GHz) to a 5 volt supply and measured the current used by each pin. The results confirm what sanoayyk has said here earlier.
Pinout:
1. V(motor), top fan
2. Tacho
3. GND (common to both fans, two wires connected)
4. Vcc (common to both fans, two wires connected)
5. V(motor), bottom fan
6. Tacho
The test configuration is the same as in this drawing by Aman Fahimullah, except he has the pin numbering the wrong way. Pin no 1 is in the same place as it is on 3 and 4-pin fan connectors. (On the right in this picture.)
V(motor) draws 79 mA of current at 5 volts. Vcc only uses 6.3 mA.
I made a similar measurement on a 120 mm case fan: it used 22 mA @ 5 volts. The 92 mm fans on the G5 use 4 times as much energy on the same voltage as the PC fan. The PC was quiet at 5 volts but the G5 fan already made distracting noise.
Conclusions:
1) Mac fans are much more powerful than any PC fans. To keep the noise down they must be operated at far lower voltages.
The control logic in brushless motors serves the function of the commutator in traditional brushed DC motors. This needs about 5 volts on Vcc to operate. PC fans typically fail to start at voltages below 5 volts. By giving the control logic a separate Vcc wire, Apple allows the fans to be operated with V(motor) voltages as low as maybe 1.5 volts. (Vcc should always be connected directly to the 12 V or 5 V rail.)
2) There is no way to trivially connect a PWM output to a G5 or early Mac Pro fan. V(motor) is not a low power "control" signals, as some have suggested. Most of the power is drawn from the V(motor) pin. (Update, May 28, 2016: Here is a
trivial circuit that converts PWM to analog voltage output.)
Also using motherboard 3-pin fan connectors may be difficult. The Mac fans may draw excessive amounts of current, especially if more than one fan is connected to the same control line. Even special 3-pin fan controllers may be ill suited, as PC fans expect voltages from 5 to 12 volts and Mac fans maybe 2 to 6 volts under normal operating conditions.
3) The simplest way to control Mac fans may be to "steal" the PWM signal from the CPU fan and use it to control all the case fans. The tachometer outputs may still be connected to different 3 or 4-pin headers on the motherboard to get RPM readings. In the simplest case there is also a normal PC-type CPU fan running to guarantee that the motherboard and BIOS get a normal tacho reading and do not start panicking about a loss of CPU cooler.
a) Connect Vcc to 12 volts, V(motor) to 5 volts and use a N-type MOSFET to switch the GND connection. The MOSFET can be directly controlled by the PWM signal.
The problem with this circuit is that the tachometer signal might not work. I guess it would be possible to direct the tacho line through a capacitor and still read it normally.
For higher speeds at 7 volts one might connect the whole fan circuit between the 5 volt and 12 volt rails. I guess it might still be possible to switch the N-type MOSFET on the 5 volt rail from the PWM output with only resistors and maybe a diode for protection. The mother board might have a Zener diode to prevent the pull-up pulling the line above 3.3 volts. See here:
http://www.pavouk.org/hw/fan/en_fan4wire.html
b) The other option is to regulate current from the 12 volt or 5 volt rail. This requires extra circuitry an a circuit board. Sanoayyk already presented one possible circuit. Too many transistors for my taste. Cannot one just use a P-type MOSFET on the 12 volt rail and an opto-isolator or something to pass through the PWM signal?
Update May 28, 2016:
There seems to be a huge misunderstanding about what the voltage control or PWM needs to do to control the Apple fans. Some have suggested that the PWM output needs to be converted to a voltage output. Others have suggested that the Mac fans can be made to spin slower by somehow lowering the voltage. None of this is relevant.
Motherboard fan control does not aim at controlling the voltage or duty cycle based on temperature. In reality the tachometer feedback is part of the control loop. The circuitry aims at certain rpm based on temperature.
Quoting from a Maxim Integrated application note:
Fan Control Advances: Consider Fan-Speed Regulation
The MAX6650 is a true fan-speed regulator because it includes a tachometer output fan in its feedback loop.
I just read somewhere that the idle speed of Apple fans is 300 rpm. This is far lower than the minimum speeds for any PC fans. For Mac fans to operate normally under Hackintosh control someone must tell the fan controller to aim for a much lower fan speed.
Motherboards may have dedicated fan control chips, like the ones made by Maxim. These are programmed and controlled over the I2C / SMBus control bus. Another quote from the same application note:
The MAX6639 monitors the tachometer outputs of two fans and adjusts the duty cycles of its PWM outputs to force the fans to the correct speeds. As with the MAX6620, the MAX6639 includes several features such as detection of fan failures and adjustable PWM ramp rate to help reduce the audibility of fan-speed changes. In addition, the MAX6639 includes 2 channels of temperature sensing and a programmable temperature-to-RPM control algorithm that enables automatic control of fan RPM as a function of temperature.
How will the operating system know how to give the fan controller the right operating parameters?