Re: [lm-sensors] small changes in configuration file for MSI

["Leo Krüger" <mail@leokrueger.de>]

Hi Jean,

> -----Ursprüngliche Nachricht-----
> Von: Jean Delvare [mailto:khali@linux-fr.org]
> Gesendet: Montag, 30. August 2010 10:24
> An: Leo Krüger
> Cc: lm-sensors@lm-sensors.org; Thomas Ettwein
> Betreff: Re: [lm-sensors] small changes in configuration file for MSI
> IM-945GSE-A
>
> Hi Leo,
>
> On Sun, 29 Aug 2010 01:16:17 +0200, Leo Krüger wrote:
> > I own a MSI IM-945GSE-A motherboard and used the sample configuration
> file
> > from http://lm-sensors.org/wiki/Configurations/MSI/IM-945GSE-A,
> > which showed me around 14V for the 12V output.
>
> I'm adding the original author of the configuration file to Cc.
>
> > Since this would be far above specification limits, I measured the
> actual
> > values at the ATX power connector. In addition to the 12V output, the
> 5V
> > and the 5Vsb line differed a little.
>
> I am a little curious about your experimental setup. How do you get
> access to the ATX power connector while the system is running?

For my atx power connector it was possible to get to conducting parts with a
small test prod.

>
> I usually measure +5V and +12V from molex connectors, it's easy. But
> 5VSB, I wouldn't know how to measure it.
>

http://en.wikipedia.org/wiki/ATX#Power_supply
the purple wire is 5vsb

> > Below is my updated configuration file. Note that I just changed 3
> lines,
> > the compute lines for in4, in5 and in6.
> > It would be nice if you could update the sample file in your wiki.
> > Let me know if any additional steps are necessary.
>
> It is expected that measuring the values with a multimeter will lead to
> slightly different results for voltages which need scaling. Scaling is
> achieved using two resistors, the value of which is typically
> guaranteed to 1%. Therefore, there can be up to 2% of error introduced
> by the resistors themselves. By calibrating your configuration file
> based on actual measurement, you correct this error for your specific
> motherboard, but the same correction doesn't apply to other instances
> of this motherboard model.
>
> So your proposed changes may be perfect for your board, but aren't
> suitable for the default configuration file for this board.
>
> A better way to figure out which scaling factors are intended by the
> board manufacturer, is to compare with what the BIOS is printing. First
> write down all values reported by the BIOS, and also note your
> voltmeter measurement. Then do the same again with lm-sensors's
> unscaled voltage values. If the voltmeter values didn't change, then
> lm-sensors should report the same as the BIOS after scaling. If the
> voltmeter values changed (which can happen as the system load is
> slightly different) then you have to take it into account in your
> computations.

I've to admit that this sounds much more reliable than my simple
measurements.
To check your proposed factors I'm going to do it the way you suggested.
Afterwards I will give you an acknowledgement.
I've access to two of these motherboards, so I will check if the results
differ.

>
> The 5.25 scaling factor comes straight from the datasheet (suggested
> resistors of 200k and 47k), and your own measurement for +5V leads to
> almost exactly this, so I wouldn't change it. It is also quite unlikely
> that the vendor would use different resistor values for +5V and 5VSB,
> which makes me reluctant to change 5VSB either, although your proposal
> is quite different there.

Ok. I think you're right. The difference is quite marginal.

>
> For +12V, the 12.83 scaling factor was apparently copied from our old
> default configuration file, which for this chip was written for an
> Epox board. The datasheet instead suggests to use resistors of 200k and
> 20k, leading to a scaling factor of 11. Note that the difference
> between this 11 and your 10.86 is less than 2%, so it is likely that
> MSI followed to the datasheet. So I would go with 11.
>

I just found the datasheet
http://www.fintek.com.tw/files/productfiles/F71882_ab.pdf
To which one do you refer?

Please let me know if I can do anything else.
Especially if there is anything I can do to provide you with hardware
specific data
to support development of lm-sensors (I actually don't know if there is
support missing for any of the hardware parts)

Regards,
Leo

> --
> Jean Delvare
> http://khali.linux-fr.org/wishlist.html

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors