lm75 compute

lm75 compute

[Mark Studebaker]

who made the default compute for lm75 temperature be divide by 2
(attributed to the Asus TX97-E)?!?!
Ticket 1423 complains.
I can't tell from the CVS log for sensors.conf.eg how it happened.
Or did this get in by mistake?
Let's not change a years-old default for the sake of one motherboard!!!

lm75 compute

[Jean Delvare]

> ----------------------------
> revision 1.77
> date: 2002/11/08 12:06:53;  author: khali;  state: Exp;  lines: +119
> -94(Khali) Add section for lm75 in sensors.conf.eg
>         Complete the lm78 section in sensors.conf.eg
> ----------------------------

BTW, note that the change was almost one year old and this is the first
person complaining AFAIK. It might be that the LM75 is so old that it's
not used anymore. But it could as well mean that I am not the only one
needing that scaling.

> I wouldn't have changed an existing configuration, but it happens that
> the LM75 had no section at that time, so I simply made one based on
> the only board I had with a LM75 on it - an Asus TX97-E.
> 
> Now I admit that what's said on ticket 1423 is plain right. The LM75
> cannot be configured in any way, so I hardly can understand how the
> reported temperature could require scaling. On the other hand, which
> chip would it be on my TX97-E, if not a LM75? There's also a LM78 on
> that board, so it sound rather logical that the other monitoring chip
> would be from National Semiconductor too.
> 
> I'll open the case and take a look, in case I'm able to see the
> monitoring chips.

I did that and could see the chips. There *is* a real LM75 right under
the CPU socket. My guess is that the LM75 is not designed to accurately
measure a CPU's temperature, so Asus placed it in such a location that
multiplying the original value by two would give a correct approximation
of the CPU's temperature. This means that I should not trust the
reported value too much - although it *is* related to how much heat the
CPU generates, whatever the actual value is.

Whatever, the LM75's value *needs* to be scaled by 2 on my system to
match what the BIOS screen reports. This is why I changed the
configuration file the way I did. Sorry for the trouble, had I read the
LM75 datasheet before doing that, I'd have understand it wasn't making
sense at all.

-- 
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/

lm75 compute

[Jean Delvare]

> who made the default compute for lm75 temperature be divide by 2
> (attributed to the Asus TX97-E)?!?!

Me.

> Ticket 1423 complains.

Interesting.

> I can't tell from the CVS log for sensors.conf.eg how it happened.

----------------------------
revision 1.77
date: 2002/11/08 12:06:53;  author: khali;  state: Exp;  lines: +119 -94
(Khali) Add section for lm75 in sensors.conf.eg
        Complete the lm78 section in sensors.conf.eg
----------------------------

> Or did this get in by mistake?

No, I made this on purpose.

> Let's not change a years-old default for the sake of one
> motherboard!!!

I wouldn't have changed an existing configuration, but it happens that
the LM75 had no section at that time, so I simply made one based on the
only board I had with a LM75 on it - an Asus TX97-E.

Now I admit that what's said on ticket 1423 is plain right. The LM75
cannot be configured in any way, so I hardly can understand how the
reported temperature could require scaling. On the other hand, which
chip would it be on my TX97-E, if not a LM75? There's also a LM78 on
that board, so it sound rather logical that the other monitoring chip
would be from National Semiconductor too.

I'll open the case and take a look, in case I'm able to see the
monitoring chips.

-- 
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/

lm75 compute

[Denis Leroy]

Salut Jean,

Yes, i think you must be right there, i can't come up with any other
explanation of why your board BIOS would scale up the temperature. The
LM75 is an air-flow temperature sensor, not a die temperature sensor. 
My point was: you're right to modify your sensors.conf to scale it up
by 2.0, but it probably shouldn't be the default setting.

As far as i know (which is not that much as i'm a software guy), the
National LM75 is very commonly used, though i could ask one of our
hardware gurus if you're interested. One of our boards has an LM75, an
LM87 and a NE1617 very close together. All three have internal temp
sensors, and all three report pretty much the exact same temperature
(typically around 25 to 30C in our strongly air-conditionned lab). The
LM87 has pins for external sensor diodes which are connected to FPGA
diodes. Those are die temperatures, and are always a lot higher
(depending of whether the FPGA has a heat-sink or not), around 40 to
50C (hyst is typically 75C, limit 85C).

I was hoping to contribute some drivers to your project, based on your
interest and if i get some time (Philips ne1617, Vitesse VSC055, or
more fun, XFP laser 10G transceivers...).

-denis

--- Jean Delvare <khali@linux-fr.org> wrote:
> I did that and could see the chips. There *is* a real LM75 right
> under
> the CPU socket. My guess is that the LM75 is not designed to
> accurately
> measure a CPU's temperature, so Asus placed it in such a location
> that
> multiplying the original value by two would give a correct
> approximation
> of the CPU's temperature. This means that I should not trust the
> reported value too much - although it *is* related to how much heat
> the
> CPU generates, whatever the actual value is.


__________________________________
Do you Yahoo!?
Exclusive Video Premiere - Britney Spears
http://launch.yahoo.com/promos/britneyspears/

lm75 compute

[Jean Delvare]

> I was hoping to contribute some drivers to your project, based on your
> interest and if i get some time (Philips ne1617, Vitesse VSC055, or
> more fun, XFP laser 10G transceivers...).

New drivers are welcome. Take a look at the doc directory for
guidelines, ask your questions on the list, we'll try to help. When you
have something working, let us know and we'll commit it to our CVS
repository (or create an account for you if you feel like commiting the
drivers by yourself).

-- 
Jean Delvare
http://www.ensicaen.ismra.fr/~delvare/

lm75 compute

[Mark Studebaker]

Jean Delvare wrote:
> 
> > ----------------------------
> > revision 1.77
> > date: 2002/11/08 12:06:53;  author: khali;  state: Exp;  lines: +119
> > -94(Khali) Add section for lm75 in sensors.conf.eg
> >         Complete the lm78 section in sensors.conf.eg
> > ----------------------------
> 
> BTW, note that the change was almost one year old and this is the first
> person complaining AFAIK. It might be that the LM75 is so old that it's
> not used anymore. But it could as well mean that I am not the only one
> needing that scaling.
> 
>

I couldn't find the change because I assumed it was so bad that it must
have been recent... I was wrong.
Thanks for researching how it happened.
Agree that individual LM75's on mobos isn't that common anymore;
which is why it wasn't caught sooner.
mds