I second aigomorla's assessment of fan controllers. The one purpose that a decent electronics circuit can serve here is to drop 12 volts down to a lower value without shedding the entire difference as heat. That takes a well-designed transistor circuit that can handle the amps in question. A rheostat, or diodes, or linear regulators, just dump the difference as heat.
This can nevertheless be a small number of watts. For example, if one has four fans totaling 1 amp at 12 volts, they're doing less work at lower voltages, and they'll roughly draw 0.75 amps at 9 volts (75%

. A basic circuit that dumps the other 3 volts as heat is producing 3 volts * 0.75 amps = 2.25 watts of heat. This isn't much, and a better circuit can only save part of this. This calculation is back-of-the-envelope, using first-order rules of thumb; one can measure exactly using an inexpensive multimeter and a custom cable to insert the multimeter in the current path.
Saving this nightlight worth of heat inside the case doesn't seem worth putting up with the crap that passes for fan controllers these days. The added complexity is just something else that can break, damaging more expensive system components. I just lost an SSD to a cheap drive bay device, so this issue is real to me.
I'm a big fan of the
Lamptron Fan-Atic: 5-port bay device; I have them in two builds now. They're just passive switches, and they're built like brick Pentiums.
As designed, this provides five 3-way toggle switches with red/blue indicator LEDs, powering five fan headers at 5 volts, OFF, or 12 volts. The 5 volts is drawn from the power supply's 5 volt source, so there is no heat shedding. In practice, if one has enough case ventilation, one leaves all switches at 5 volts.
For example, I swapped in two Cooler Master Blade Master PWM fans, controlled by the motherboard CPU fan header (*), for my Noctua NH-D14 cooler, scaled in my ASUS P8P67 Pro BIOS to 10% at 57 C, 100% at 75 C. My case is a Mountain Mods Plateau, and I leave my nine Scythe Scythe S-Flex SFF21E case fans at 5 volts. At a modest Core-i7 2600K overclock of 4.4 GHz, my core temps are 61 C to 63 C at full load, driving the PWM fans at 30% to 40%. My system is close to silent at 4.4 GHz full load, a great compromise over the louder 4.8 GHz overclock (par for this cpu).
So with Sandy Bridge,
any fan controller is like renting a storeroom rather than throwing out junk. With PWM cooler fans and 5 volt case fans, fan controllers aren't really needed, except for extreme overclocks. Even then, with ample case ventilation the case fan voltage makes a scant difference in core temps, not worth the added noise. And knobs can be rotated down to the point where fans stall; my fans don't stall at 5 volts. Nevertheless, the toggles and LEDs are nice case bling, allowing me to confirm my cooling assertions when desired. Fan cables can't handle the current of many fans, so the Fan-Atic serves as a convenient power splitter while looking decorative, avoiding a jungle of Y-splitters inside the case.
As an abstract DIY component, the Lamptron Fan-Atic comes into its own. It inputs
whatever power sources one puts on the "5 V" and "12 V" pairs of lines coming into the 4-pin power header. The toggles then select, fan by fan, which source to use. For example, one could wire a string of diodes or a beefy rheostat into the 12 volt source, leaving the 5 volt source alone. I have a string of 3 amp diodes feeding a Q6600 build Fan-Atic, allowing me to choose between 5 volts and 8 volts. If one is comfortable with the risk to one's power supply, one could also rewire the 12 volt source using the infamous "7 volt trick".
Lamptron Fan-Atic: 5-port
(Sidewinder) (Silicon Valley Compucycle) (Xoxide) (Amazon) (Performance PC's)
Sidewinder 50 Watt Rheostat
(*) To state the obvious, a PWM Y-splitter cable should only return one RPM signal. Various sources sell splitters wired in the "obvious way", hooking up all wires where they look like they should go. This is stupid, but easily fixed by playing "Mission Impossible" background music and deciding which wire to snip. The unused pin then pulls out easily.