Basically if you got 1 or 2 fans you need to power up you can usually just connect it to the motherboard and control the fan speed that way.
However if you got a server chassis and you got many fans and you don't want them running at full speed due to noise, then you will need to regulate the speed somehow.
Depending on which fans you have, you can wire them up in series instead of parallel so 2 fans will use +6V instead of +12V and it might or might not be the fan speed you are looking for. Keep in mind for this to work they need to be identical fans.
You can also get a variable resistor and just control it that way, just make sure it can handle the amperage if you are using industrial type fans since the mini variable knobs might not be able to handle it and end up melting and cause a fire.
^^^ This, this is what I was wondering.
Undervolting fans, running them in series, etc... is very wasteful electricity wise compared to pulse width modulation. I've run into a similar problem with my gaming rig because I run three san ace 120 fans (all PWM, 4 pin fans, 4 amps each, 48 watts at full blast.) For now I'm using a controller I designed myself in 2012 to control the fan connected to the radiator for my CPU and a signal generator to temporarily control the secondary fans that are hooked into my GPU.
I put my GTX 1060 mini on an AIO water cooler using a Kraken G12 adapter and now see temps of 34C under full load, overclocked 150mhz on the gpu. I haven't touched the VRAM overclock yet because I need to add additional cooling.
I may go back and develop a controller that allows for independent PWN fan control that interfaces with a piece of software in windows/linux.
My understanding is profit margin is tied to electricity consumption, yes? Do you guys think there might be any demand for such a device? It would reduce energy waste from the cooling and not generate any heat, unlike rheostats, voltage regulators, etc... It would also be easier on the fans themselves. Less wear and tear.
If so I may design the tool and maybe put up a kickstarter. I have all the tools necessary, but prototyping pcb's and getting the cheapest copy of Proteus for VSM simulation would set me back about $200-$300. If not I suppose I'll still develop it, but it will take me longer without a copy of Proteus. Breadboarding, programming, and debugging in real life is a lot slower than what can be done in a VSM simulator!