On Mon, Jan 5, 2009 at 6:41 PM, John R Pierce <pierce@xxxxxxxxxxxx> wrote:
> Joseph L. Casale wrote:
<snip>
>> Yeah, turn on a machine w/o a heatsink :) In a couple of seconds it will
>> get so hot that you can't touch it and the internal safety threshold on the
>> proc will shut it down. I suspect the mobo might have a bios that can control
>> the fan speed and the hysteresis or whatever technique it uses to control
>> reaction time. My Asus has this IIRC...
>
> the newest Intel chipsets actually have all this built into the ICH9R or
> whatever, managed by microcode running in the northbridge as a
> management coprocessor. I forget what Intel 4-letter-acronym this is
> under. However, this does NOT play well with the legacy LM style way
> of doing things from the host, and I suspect many of the third party
> boards continue to use LM or similar sensor chips on the i2c bus, old
> school, rather than relying on this chipset built-in fan control
> stuff. I have an Intel branded G33 board (running Windows XP) and
> none of the usual fan control programs like SpeedFan have a clue how to
> talk to this stuff. I'll guess without any specific knowlege that
> Linux lm_sensors will have similar problems.
>
> Intel calls this coprocessor on the Northbridge the Management Engine
> (ME) and Host Embedded Controller Interface (HECI), and system wide its
> branded Intel "Quiet System Technology" (QST), which includes the MCH w/
> ME, the ME Firmware, and the associated BIOS support. Apparently once
> configured by the BIOS, this runs without host OS intervention,
> including monitoring the CPU, northbridge and southbridge thermal
> sensors, cpu and chassis fan tach sensors and PWM for the respective fans.
The HW people are way ahead of the SW people? A good reason not to use
the latest motherboards, with this technology, in production servers,
at this time?
_______________________________________________
CentOS mailing list
CentOS@xxxxxxxxxx
http://lists.centos.org/mailman/listinfo/centos
