[prev in list] [next in list] [prev in thread] [next in thread] 

List:       lm-sensors
Subject:    Re: [lm-sensors] high value for CPUTIN
From:       Guenter Roeck <linux () roeck-us ! net>
Date:       2012-10-31 3:58:45
Message-ID: 20121031035845.GA25948 () roeck-us ! net
[Download RAW message or body]

On Tue, Oct 30, 2012 at 11:18:00PM -0300, Alejandro Jakubi wrote:
> >No. Problem is that PECI does not report an absolute temperature, but the
> >difference to Tjmax, which is the maximum CPU temperature (reported as critical
> >temperature by coretemp). The NCT6776F has a register which needs to be programmed
> >to that value. Usually that is done in the BIOS. Looks like the BIOS did not
> >program the correct value on your board.
> 
> I see. However, if this NCT6776F register were wrongly programed, it seems
> to me like there should be a systematic error in the reported temperature.
> However, this is not what I am observing. For instance, the value of PECI 0 in
> my first post (showing the complete output) was 33.5C, with all the other
> temperatures (except CPUTIN) in a close range. Now its value is about 30C (+/-
> 1C, I am following it with xsensors), again in close agreement with the other
> readings.
> 
> So, it seems to me like something else caused yesterday that PECI 0 had
> reported a value over 10C below "normal". Two things come to my mind: an
> erratic failure of this device, or that previously to the reported low
> reading, I had booted to Win 7 and installed SpeedFan 4.47 (as a check about
> the high value reported for the CPU).
> 
> I will keep looking at the PECI 0 value as summer comes here shortly...
> 
PECI is actually _very_ inaccurate for low temperatures, so this is not really
surprising.

> In any case, are the values reported by Core 0 to 3 more reliable?
> 
Good question. I don't really know; those are also known to be inaccurate
for low temperatures.

Guenter

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors
[prev in list] [next in list] [prev in thread] [next in thread]