On 10/12/2014 10:15 AM, Phil Pokorny wrote:
On Sun, Oct 12, 2014 at 9:12 AM, Felix Schulthess <fsch@xxxxxxxxxxx> wrote:
You cannot control the fan speeds. They are managed by the onboard
controller. You can choose one of several profiles in the bios but that
is it.
That explains a lot. Thank you for your help, I am going to do some
further research in this direction and read up on FreeIPMI.
Still, if I can't control the fan speeds but they are instead controlled
by some onboard controller, then how could the installation on the
fancontrol and lm-sensors mess up the control loop so bad? This still
keeps me wondering.
You can't control the fan speeds because there is already another device on
the motherboard that is tasked with the job of controlling them. As you
found, you can poke the control chips, but that doesn't mean you can
control the speeds predictably. Imagine two cooks in a kitchen trying to
cook different meals with the same ingredients. They come and go in the
kitchen unaware of the others meddling and confused why things are changing
when they are away. You end up with no food. Only one cook in the kitchen.
It's also possible that when you ran SDT it's default config was to change
the BIOS/BMC setting on the fans speeds to full speed. As you found it was
a permanent change to the setting, not some strange interaction between
lm_sensors and sdt and the BMC.
Maybe, the sensors somehow blocked the CPU temperature readout. This
could have caused the superdoctor program to resort to some failsafe
behavior (i.e. spinning the fans up to max RPM). Just wondering.
Also possible.
No idea how that could happen, unless there is a means for an application
or kernel module to block access to specific CPU registers. That would be
news to me, though.
Guenter
_______________________________________________
lm-sensors mailing list
lm-sensors@xxxxxxxxxxxxxx
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
