[lm-sensors] does lm-sensors pick up sensors that don't exist?

[lm-sensors] does lm-sensors pick up sensors that don't exist?

[Per Jessen]

Using lm-sensors I've been getting 3 temperature readings from a newly
bought system - system, CPU and an unidentified temp3.  I've been
suspecting temp3 of being the Northbridge temperature, but I've just
been told by Gigabyte support that there is no sensor for temp3 ... 

Is this a "normal" thing to see?  I've been particularly interested in
this temp3 as it has been hovering in the 80-86 degrees range, which I
thought might be the cause of severe system instability (whilst under
load). 


/Per Jessen, Zürich


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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Matt Roberds]

On Mon, 18 Feb 2008, Per Jessen wrote:
> I've been suspecting temp3 of being the Northbridge temperature, but
> I've just been told by Gigabyte support that there is no sensor for
> temp3 ...

Many monitoring chips have multiple inputs and the motherboard
manufacturer does not always connect all of them.  I have an Asus
M2N-SLI motherboard with an it8716 monitoring chip with three inputs,
but only two of them are connected.

Another possibility is that temp3 is connected to *something* in your
system, but the values that lm-sensors uses to scale the raw reading
are incorrect.  Have you looked in the /etc/sensors.conf or
/etc/sensors3.conf for any information on your particular monitoring
chip or motherboard?

> I've been particularly interested in this temp3 as it has been
> hovering in the 80-86 degrees range, which I thought might be the
> cause of severe system instability (whilst under load).

On my system, temp3 on the it8716 always reports the maximum possible
value, +127 C.  I can guarantee (by putting my hand inside the case)
that nothing in my computer is that hot, so I have always just ignored
that reading.  I suspect that either the chip is designed to report the
maximum possible reading if there is no sensor connected, or that Asus
wired the unused sensor input to ground, +3.3 V, or some other voltage
that makes it read at the maximum possible.

Matt Roberds


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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Jean Delvare]

Hi Matt,

On Mon, 18 Feb 2008 08:30:09 -0600 (CST), Matt Roberds wrote:
> On my system, temp3 on the it8716 always reports the maximum possible
> value, +127 C.  I can guarantee (by putting my hand inside the case)
> that nothing in my computer is that hot, so I have always just ignored
> that reading.  I suspect that either the chip is designed to report the
> maximum possible reading if there is no sensor connected, or that Asus
> wired the unused sensor input to ground, +3.3 V, or some other voltage
> that makes it read at the maximum possible.

Depends on the thermal sensor type. For thermal diodes (or
diode-connected transistors) the chip can typically detect if the
thermal sensor is missing, and it will report it either explicitly in a
status register, or through an arbitrary value (typically -128 or +127).

For thermistors, what the chip measures is actually a voltage, which is
then converted to a temperature value. If the board manufacturer
doesn't want to implement a sensor, they will typically wire the input
to the ground, which is equivalent to an infinitely high or infinitely
low temperature (depending on how the voltage divisor bridge is built),
and the chip doesn't have to treat this as a special case. Things get
bad when the manufacturer leave the thermal input floating, you will
get random temperatures. This is quite possibly what Per is
experiencing.

-- 
Jean Delvare

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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Per Jessen]

Jean Delvare wrote:

> Depends on the thermal sensor type. For thermal diodes (or
> diode-connected transistors) the chip can typically detect if the
> thermal sensor is missing, and it will report it either explicitly in
> a status register, or through an arbitrary value (typically -128 or
> +127).
> 
> For thermistors, what the chip measures is actually a voltage, which
> is then converted to a temperature value. If the board manufacturer
> doesn't want to implement a sensor, they will typically wire the input
> to the ground, which is equivalent to an infinitely high or infinitely
> low temperature (depending on how the voltage divisor bridge is
> built), and the chip doesn't have to treat this as a special case.
> Things get bad when the manufacturer leave the thermal input floating,
> you will get random temperatures. This is quite possibly what Per is
> experiencing.

Hi Jean

if only I was getting random readings, but the readout I'm seeing
doesn't look random at all - it's typically 80-81, but will increase to
86 when I'm stressing the system (the CPU-temp will rise to 62/63 at
the same time).  

I'm currently waiting for the board to be replaced, so I can't tell you
exactly what lm-sensors says about the type of sensor, but I think it
said it was "transistor". 

Regardless, I guess I should be ignoring it when Gigabyte says it's not
there ...



/Per Jessen, Zürich


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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Jean Delvare]

On Mon, 18 Feb 2008 16:20:21 +0100, Per Jessen wrote:
> Jean Delvare wrote:
> 
> > Depends on the thermal sensor type. For thermal diodes (or
> > diode-connected transistors) the chip can typically detect if the
> > thermal sensor is missing, and it will report it either explicitly in
> > a status register, or through an arbitrary value (typically -128 or
> > +127).
> > 
> > For thermistors, what the chip measures is actually a voltage, which
> > is then converted to a temperature value. If the board manufacturer
> > doesn't want to implement a sensor, they will typically wire the input
> > to the ground, which is equivalent to an infinitely high or infinitely
> > low temperature (depending on how the voltage divisor bridge is
> > built), and the chip doesn't have to treat this as a special case.
> > Things get bad when the manufacturer leave the thermal input floating,
> > you will get random temperatures. This is quite possibly what Per is
> > experiencing.
> 
> Hi Jean
> 
> if only I was getting random readings, but the readout I'm seeing
> doesn't look random at all - it's typically 80-81, but will increase to
> 86 when I'm stressing the system (the CPU-temp will rise to 62/63 at
> the same time).  

Well, if it seems to make some sense, it might as well be wired.
However, 80°C is rather high, and I would hope that no chip on my
motherboard gets this hot. But depending on the hardware, it might
actually happen.

> I'm currently waiting for the board to be replaced, so I can't tell you
> exactly what lm-sensors says about the type of sensor, but I think it
> said it was "transistor". 

BTW, you might try switching the sensor type to see if it gives more
sensible results. In theory the BIOS should set the right sensor types
for you, but it doesn't always do so.

> Regardless, I guess I should be ignoring it when Gigabyte says it's not
> there ...

You mean you would trust what the support of a motherboard vendor told
you? Naaaah ;) The sad truth is that most of the time they have no idea
what you are talking about, and come up with the reply that is the most
likely to make you go away.

-- 
Jean Delvare

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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Mark M. Hoffman]

Hi:

* Jean Delvare <khali@linux-fr.org> [2008-02-18 16:38:58 +0100]:
> On Mon, 18 Feb 2008 16:20:21 +0100, Per Jessen wrote:
> > Jean Delvare wrote:
> > 
> > > Depends on the thermal sensor type. For thermal diodes (or
> > > diode-connected transistors) the chip can typically detect if the
> > > thermal sensor is missing, and it will report it either explicitly in
> > > a status register, or through an arbitrary value (typically -128 or
> > > +127).
> > > 
> > > For thermistors, what the chip measures is actually a voltage, which
> > > is then converted to a temperature value. If the board manufacturer
> > > doesn't want to implement a sensor, they will typically wire the input
> > > to the ground, which is equivalent to an infinitely high or infinitely
> > > low temperature (depending on how the voltage divisor bridge is
> > > built), and the chip doesn't have to treat this as a special case.
> > > Things get bad when the manufacturer leave the thermal input floating,
> > > you will get random temperatures. This is quite possibly what Per is
> > > experiencing.
> > 
> > Hi Jean
> > 
> > if only I was getting random readings, but the readout I'm seeing
> > doesn't look random at all - it's typically 80-81, but will increase to
> > 86 when I'm stressing the system (the CPU-temp will rise to 62/63 at
> > the same time).  
> 
> Well, if it seems to make some sense, it might as well be wired.
> However, 80°C is rather high, and I would hope that no chip on my
> motherboard gets this hot. But depending on the hardware, it might
> actually happen.

Note: "floating" doesn't necessarily imply "random".  A floating input
could still show some small correlation with other circuits in close
physical proximity.

Regards,

-- 
Mark M. Hoffman
mhoffman@lightlink.com


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

Re: [lm-sensors] does lm-sensors pick up sensors that don't exist?

[Jean Delvare]

Hi Mark,

On Mon, 18 Feb 2008 22:21:07 -0500, Mark M. Hoffman wrote:
> * Jean Delvare <khali@linux-fr.org> [2008-02-18 16:38:58 +0100]:
> > Well, if it seems to make some sense, it might as well be wired.
> > However, 80°C is rather high, and I would hope that no chip on my
> > motherboard gets this hot. But depending on the hardware, it might
> > actually happen.
> 
> Note: "floating" doesn't necessarily imply "random".  A floating input
> could still show some small correlation with other circuits in close
> physical proximity.

I suspected this but didn't know for sure - thanks for the confirmation!

-- 
Jean Delvare

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