preferred fan speed control methods?

[Ben Scott]

On 11/29/06, Bill McGonigle <bill at bfccomputing.com> wrote:
> I've got a couple very noisy servers that are well under operating range for their
> components and I'd like to slow down the fans a bit...

  Keep in mind that, in general, the cooler a component is, the longer
it lasts.  So running well under operating range can still yield
benefits.  Maximum != ideal.

  That being said, choice of fan control is largely dependent on the hardware.

  If the fans are wired right to power supply lines, software control
is out (without something expensive).  But you can get manual speed
control knobs for cheap, and simple auto-controls with a temperature
sensors for not too much more.

  If the fans connect to the mainboard, it depends on the mainboard.
Some mainboard fan headers are just power supply taps.  No control
circuits.  See above.  Be aware that some boards can monitor fan
speed, but not control it, so the presence of a speed monitor isn't a
guarantee there's a control circuit.

  If the mainboard has speed control capability, then it depends on
how it's done.  Sometimes the speed control is hard-wired to a
thermistor or whatever, with no software control possible.  Again, see
above.

  If you do have honest-to-goodness software control, then different
chipsets and boards still use different mechanisms. There's little to
no standardization for this.

  I briefly played around with some of this on my Dell Precision 380.
The machine does have *some* kind of fthermal management (monitoring,
at least).  The "lmsensors" package found the system management bus
controller, but didn't identify anything useful on it.  I gave up
quickly, I'm afraid.

-- Ben