[lm-sensors] Biostar TA880GB+ Motherboard Information

[lm-sensors] Biostar TA880GB+ Motherboard Information

[Paul Norman]

I was going to add this to the wiki, but was unable to figure out how to
register to add it. Someone on IRC suggested I post to the mailing list.

My Ubuntu server runs on a Biostar TA880GB+ AM3 motherboard, with a 880G
northbridge and an Athelon II processor.

Finding no configuration information for this motherboard, I proceeded to
write it, using the fairly detailed information in the BIOS to match up
sensors with what they measure. I overvolted parts in the BIOS to verify
that I had the configuration right.

My configuration file is as follows. I am running sensors version 3.3.0 with
libsensors version 3.3.0

chip "it8721-isa-*"

# Temperatures
    label temp1 "CPU"
    label temp2 "Northbridge"
    label temp3 "Mainboard"

# fans
# Assuming fans are connected as specified on the motherboard.
# CPU fan is a 4 pin PWM connector, the others are 3 pin.
    label fan1 "CPU Fan"
    label fan2 "System Fan 1"
    label fan3 "System Fan 2"

# Voltages
    #Not sure what in0 and in1 are. They're stable, maybe +12V and +5V lines
in need of scaling?
    label in2 "Vcore"
    label in3 "+3.3V"
    label in4 "NB Voltage"
    label in5 "DDR Voltage"
    label in6 "HT Voltage"
    label in7 "+3.3Vsb"


chip "acpitz-virtual-0"
# This is essentially the same as the CPU temperature.
    label temp1 "temp1"

chip "k10temp-pci-00c3"
#    temp1 represents the CPU temperature but is not in any physical unit
    label temp1 "temp2"

The Northbridge temperature is not listed in the BIOS, but it's the only
component that runs that hot, is not affected by CPU load, and is affected
by CPU fan speed (the cooler gets some air to the northbridge)

I'm pretty sure that in0 and in1 are +5V and +12V readings since those are
shown in the BIOS, but I don't know of a way to tell them apart.

Fancontrol does work with this setup although with my cooling it doesn't
accomplish much since the loudest and most powerful fans are cooling my
drives. Biostar makes some other 880G-based motherboards and this
configuration may apply to them too.

-- 
Paul Norman
http://www.paulnorman.ca/blog


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

Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

[Jean Delvare]

Hi Paul,

On Fri, 18 Nov 2011 22:32:09 -0800, Paul Norman wrote:
> I was going to add this to the wiki, but was unable to figure out how to
> register to add it. Someone on IRC suggested I post to the mailing list.

Yes, the wiki is read-only for user to avoid spam.

> My Ubuntu server runs on a Biostar TA880GB+ AM3 motherboard, with a 880G
> northbridge and an Athelon II processor.
> 
> Finding no configuration information for this motherboard, I proceeded to
> write it, using the fairly detailed information in the BIOS to match up
> sensors with what they measure. I overvolted parts in the BIOS to verify
> that I had the configuration right.
> 
> My configuration file is as follows. I am running sensors version 3.3.0 with
> libsensors version 3.3.0
> 
> chip "it8721-isa-*"
> 
> # Temperatures
>     label temp1 "CPU"
>     label temp2 "Northbridge"
>     label temp3 "Mainboard"
> 
> # fans
> # Assuming fans are connected as specified on the motherboard.
> # CPU fan is a 4 pin PWM connector, the others are 3 pin.
>     label fan1 "CPU Fan"
>     label fan2 "System Fan 1"
>     label fan3 "System Fan 2"
> 
> # Voltages
>     #Not sure what in0 and in1 are. They're stable, maybe +12V and +5V lines
> in need of scaling?
>     label in2 "Vcore"
>     label in3 "+3.3V"
>     label in4 "NB Voltage"
>     label in5 "DDR Voltage"
>     label in6 "HT Voltage"
>     label in7 "+3.3Vsb"
> 
> 
> chip "acpitz-virtual-0"
> # This is essentially the same as the CPU temperature.

This could be problematic. If ACPI reads from the same registers as the
IT8712F chip, as these accesses are not synchronized, they could
collide. This is a problem affecting many boards unfortunately, with no
good solution so far.

>     label temp1 "temp1"

This is a no-op.

 chip "k10temp-pci-00c3"
> #    temp1 represents the CPU temperature but is not in any physical unit
>     label temp1 "temp2"

This is pretty confusing. Either leave it alone or give it a proper
name ("CPU").

The fact that the value doesn't seem to match the reality is known.
Some CPU models have broken sensors. The rest returns a relative value.
From Documentation/hwmon/k10temp:

"There is one temperature measurement value, available as temp1_input in
sysfs. It is measured in degrees Celsius with a resolution of 1/8th degree.
Please note that it is defined as a relative value; to quote the AMD manual:

  Tctl is the processor temperature control value, used by the platform to
  control cooling systems. Tctl is a non-physical temperature on an
  arbitrary scale measured in degrees. It does _not_ represent an actual
  physical temperature like die or case temperature. Instead, it specifies
  the processor temperature relative to the point at which the system must
  supply the maximum cooling for the processor's specified maximum case
  temperature and maximum thermal power dissipation."

Unfortunately we still lack a way to let user-space know about this
limitation.

> The Northbridge temperature is not listed in the BIOS, but it's the only
> component that runs that hot, is not affected by CPU load, and is affected
> by CPU fan speed (the cooler gets some air to the northbridge)
> 
> I'm pretty sure that in0 and in1 are +5V and +12V readings since those are
> shown in the BIOS, but I don't know of a way to tell them apart.

Go to the BIOS and watch the +5V and +12V values. Write down every
different sample you manage to get. The more samples, the better. Then
check the raw values (from sensors -u) for in0 and in1. Write down all
samples as well if there are many. From that we can find out who is who
and the scaling factors.

> Fancontrol does work with this setup although with my cooling it doesn't
> accomplish much since the loudest and most powerful fans are cooling my
> drives. Biostar makes some other 880G-based motherboards and this
> configuration may apply to them too.

-- 
Jean Delvare

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

Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

[Paul Norman]

> From: Jean Delvare [mailto:khali@linux-fr.org]
> To: Paul Norman
> Subject: Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

> > chip "acpitz-virtual-0"
> > # This is essentially the same as the CPU temperature.
> 
> This could be problematic. If ACPI reads from the same registers as the
> IT8712F chip, as these accesses are not synchronized, they could
> collide. This is a problem affecting many boards unfortunately, with no
> good solution so far.

Would disabling it be a good idea then?

Also, do you have any suggestions for the k10temp-pci-00c3 temp1 label?

> Go to the BIOS and watch the +5V and +12V values. Write down every
> different sample you manage to get. The more samples, the better. Then
> check the raw values (from sensors -u) for in0 and in1. Write down all
> samples as well if there are many. From that we can find out who is who
> and the scaling factors.

According to the BIOS +12V is at 12.250 with no variation and +5V is at
5.115. I waited for about 2 minutes.

Both in0_input and in1_input vary between 2.940 and 2.952. They vary
independently, and I only observed in0_input=2.940 under heavy CPU load.

I only observed in1_input=2.940 under heavy IO load.



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

Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

[Jean Delvare]

On Sat, 19 Nov 2011 02:18:49 -0800, Paul Norman wrote:
> > From: Jean Delvare [mailto:khali@linux-fr.org]
> > To: Paul Norman
> > Subject: Re: [lm-sensors] Biostar TA880GB+ Motherboard Information
> 
> > > chip "acpitz-virtual-0"
> > > # This is essentially the same as the CPU temperature.
> > 
> > This could be problematic. If ACPI reads from the same registers as the
> > IT8712F chip, as these accesses are not synchronized, they could
> > collide. This is a problem affecting many boards unfortunately, with no
> > good solution so far.
> 
> Would disabling it be a good idea then?

If the thermal zone is only used by your system for reporting, yes. If
it is also used by the BIOS to achieve thermal protection, then you'd
lose this feature by disabling the thermal zone. That being said, using
both it87 and the ACPI thermal zone is worse than any alternative, so
if you really want to monitor your system, disabling the ACPI thermal
zone is the way to go. That's what I would do personally.

> Also, do you have any suggestions for the k10temp-pci-00c3 temp1 label?

It is definitely related to the CPU temperature, even though the
numbers are not physical. If you are happy with the CPU temperature
reported by the it87 driver, you could blacklist the k10temp driver.

> > Go to the BIOS and watch the +5V and +12V values. Write down every
> > different sample you manage to get. The more samples, the better. Then
> > check the raw values (from sensors -u) for in0 and in1. Write down all
> > samples as well if there are many. From that we can find out who is who
> > and the scaling factors.
> 
> According to the BIOS +12V is at 12.250 with no variation and +5V is at
> 5.115. I waited for about 2 minutes.

Note that some BIOS do not refresh the monitored value in real-time. In
that case you have to leave and re-enter the monitoring panel to get a
chance to see different values.

> Both in0_input and in1_input vary between 2.940 and 2.952. They vary
> independently, and I only observed in0_input=2.940 under heavy CPU load.
> 
> I only observed in1_input=2.940 under heavy IO load.

Hu. If both have exactly the same raw value, it's obviously difficult
to distinguish between them. And the only other Biostar configuration
file we have has a completely different mapping, so it won't help.

Your only chance would be to temporarily change your hardware setup to
slightly alter the load. For example adding or removing a hard disk
drive, or using a different PSU.

I am a little surprised by the raw values BTW... 2.952 V is really
close to the ADC max of 3.060 V.

-- 
Jean Delvare

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

Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

[Paul Norman]

> From: Jean Delvare [mailto:khali@linux-fr.org]
> Subject: Re: [lm-sensors] Biostar TA880GB+ Motherboard Information
> 
> > Both in0_input and in1_input vary between 2.940 and 2.952. They vary
> > independently, and I only observed in0_input=2.940 under heavy CPU
> load.
> >
> > I only observed in1_input=2.940 under heavy IO load.
> 
> Hu. If both have exactly the same raw value, it's obviously difficult to
> distinguish between them. And the only other Biostar configuration file
> we have has a completely different mapping, so it won't help.
> 
> Your only chance would be to temporarily change your hardware setup to
> slightly alter the load. For example adding or removing a hard disk
> drive, or using a different PSU.

I tried adding/removing a load from the 12V line composed of a cold cathode
tube and all the fans I could control in the case, totaling about 10-15W of
draw on the +12V line and none on the +5V. Removing everything brought up
the in0 value to 2.964V. Attempts to change the BIOS value were unsuccessful
as I have no control over the CPU power draw while in BIOS. I hooked up a
multimeter and I was unable to get voltage changes that caused the needle to
move.

Because adjusting the load on the 12V line changed in0 but not in1 I think
in0 is 12V and in1 is 5V.

I think the reason I can't influence the voltages much is that the power
supply is rather excessive for what it powers. It draws ~125W at the wall
and it's a 600W power supply. It's +12V rail is speced to power multiple
graphics cards, not a low power Athlon II CPU and some fans.

When I get my SATA card back from RMA I can try adding drives - right now
I'm maxed out.

> I am a little surprised by the raw values BTW... 2.952 V is really close
> to the ADC max of 3.060 V.

This surprises me too. Based on the values and an ADC max of 3.060 the
sensors would max out at a voltage of 5.8-6.1% over the nominal voltage.
Iirc, the power supply spec allows +/- 5%, so that seems to be cutting it a
bit fine. I suppose it works if your power supply is within spec, but I bet
there's lots out there more than 5% off.


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

Re: [lm-sensors] Biostar TA880GB+ Motherboard Information

[Jean Delvare]

Hi Paul,

On Sat, 19 Nov 2011 13:21:06 -0800, Paul Norman wrote:
> I tried adding/removing a load from the 12V line composed of a cold cathode
> tube and all the fans I could control in the case, totaling about 10-15W of
> draw on the +12V line and none on the +5V. Removing everything brought up
> the in0 value to 2.964V. Attempts to change the BIOS value were unsuccessful
> as I have no control over the CPU power draw while in BIOS. I hooked up a
> multimeter and I was unable to get voltage changes that caused the needle to
> move.

I recommend using digital voltmeters for this work. It doesn't matter
too much how accurate they are. What matters is that it is easier to
distinguish between two close values.

> Because adjusting the load on the 12V line changed in0 but not in1 I think
> in0 is 12V and in1 is 5V.

Yes, this makes sense. Given the values you gave previously, this would
lead to a scaling factor of about 4.1666 for +12V (25/6) and 1.7397 for
+5V (I couldn't find a rational number for it.)

> I think the reason I can't influence the voltages much is that the power
> supply is rather excessive for what it powers. It draws ~125W at the wall
> and it's a 600W power supply. It's +12V rail is speced to power multiple
> graphics cards, not a low power Athlon II CPU and some fans.

FWIW, I wouldn't recommend this. PSU are typically most efficient
around 50% of their rated load, so using a big PSU on a small system
increases power losses.

> When I get my SATA card back from RMA I can try adding drives - right now
> I'm maxed out.
> 
> > I am a little surprised by the raw values BTW... 2.952 V is really close
> > to the ADC max of 3.060 V.
> 
> This surprises me too. Based on the values and an ADC max of 3.060 the
> sensors would max out at a voltage of 5.8-6.1% over the nominal voltage.
> Iirc, the power supply spec allows +/- 5%, so that seems to be cutting it a
> bit fine. I suppose it works if your power supply is within spec, but I bet
> there's lots out there more than 5% off.

Correct. Maybe they consider that anything beyond +5% is bad enough
anyway and can be reported as a single "too much". At least, aiming at
the end of the ADC range gives good measurement resolution, so it makes
some sense. Still, I prefer the traditional "nominal value is 3/4 of
ADC scale" rule.

When you have come up with a final configuration file for your board,
please send it and I'll add it to the wiki.

-- 
Jean Delvare

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