From: Adam on
Wes Newell wrote:
>> M/B Temp: +43�C (low = +15�C, high = +40�C) sensor = thermistor
>> CPU Temp: +30�C (low = +15�C, high = +45�C) sensor = diode
>> Temp3: +0�C (low = +15�C, high = +45�C) sensor = disabled
>
> It's likely the MB and CPU temps labels are reversed, but they could be
> right. Best way to tell is load the cpu and see which one rises fast.

Thanks, Wes! I tried playing around with /etc/sensors.conf and got
varying results. After 60 seconds of "stress," the "M/B 43, CPU 30" had
risen to "M/B 43, CPU 39." Not sure whether the labels are correct or
reversed, but I'm pretty sure the IT8705, which converts the voltage and
temperature sensors to digital, is misconfigured. It also indicates
"ALARM" for voltages that are close enough. Important thing is that,
since removing flaky DIMM, there don't seem to be any
temperature-related problems with the system.

Adam
From: Adam on
Moe Trin wrote:
>> It's likely the MB and CPU temps labels are reversed, but they could be
>> right. Best way to tell is load the cpu and see which one rises fast.
>
> I'd think the labels being reversed is much less likely than the
> calibration of the sensor/measurement tool being out of wack. Yes, the
> test you propose would answer that one way or another.

With default configuration for lm_sensors, sixty seconds of "stress"
cause a rise from MB 40, CPU 30 to MB 43, CPU 39. After playing around
with sensors.conf, most plausible result so far was rise from CPU 43, MB
29 to CPU 45, MB 40.

> Yeah, but have you actually verified that the temperatures being read
> are reasonably accurate? Certainly it's a lot easier to stick a
> thermometer in between the PCI slots than to jump through the hoops
> the CPU manufacturer specifies to actually measure the die temperature.

If I get really serious about this, I'll get a digital thermometer with
remote sensor, and assume that's the most accurate reading of all.

Adam
From: Wes Newell on
On Mon, 08 Oct 2007 00:29:53 +0000, Adam wrote:

> Wes Newell wrote:
>>> M/B Temp: +43�C (low = +15�C, high = +40�C) sensor = thermistor
>>> CPU Temp: +30�C (low = +15�C, high = +45�C) sensor = diode
>>> Temp3: +0�C (low = +15�C, high = +45�C) sensor = disabled
>>
>> It's likely the MB and CPU temps labels are reversed, but they could be
>> right. Best way to tell is load the cpu and see which one rises fast.
>
> Thanks, Wes! I tried playing around with /etc/sensors.conf and got
> varying results. After 60 seconds of "stress," the "M/B 43, CPU 30" had
> risen to "M/B 43, CPU 39." Not sure whether the labels are correct or
> reversed, but I'm pretty sure the IT8705, which converts the voltage and
> temperature sensors to digital, is misconfigured. It also indicates
> "ALARM" for voltages that are close enough. Important thing is that,
> since removing flaky DIMM, there don't seem to be any
> temperature-related problems with the system.
>
That confirms that the cpu sensor is correct, which more than likely means
the MB sensor is too. For the alarm problems look at etc/sensors.conf and
check the values there. You also have to run sensors -s on boot or
manually as root for it to take affect. Run it ince as root then run
sensors again and dsee if you still get the alarms. If so, then you'll
need to edit the IT87 portion of etc/sensors.conf. Also note the many
notes in this section as to apply to whatever MB you have.



--
Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
My server http://wesnewell.no-ip.com/cpu.php
HD Tivo S3 compared http://wesnewell.no-ip.com/mythtivo.htm

From: Wes Newell on
On Mon, 08 Oct 2007 02:05:22 +0000, Adam wrote:

> I'm beginning to think that the IT8705F is misconfigured. I looked
> through the lm_sensors sensors.cfg file, and it didn't mention my m/b,
> but it mentioned other brands with the same chip that measured things
> differently. I tried some of those settings for lm_sensors, and I think
> the most plausible one went from CPU 43C, MB 30C to CPU 45C, MB 40C
> after sixty seconds of that "CPU stress" program.

There's lots of ways to determine which is which. Use your imagination.
Locate MB sensor, heat it or cool it and see what changes.

--
Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
My server http://wesnewell.no-ip.com/cpu.php
HD Tivo S3 compared http://wesnewell.no-ip.com/mythtivo.htm

From: Unruh on
Adam <look(a)bottom.for.address> writes:

>Moe Trin wrote:
>>> It's likely the MB and CPU temps labels are reversed, but they could be
>>> right. Best way to tell is load the cpu and see which one rises fast.
>>
>> I'd think the labels being reversed is much less likely than the
>> calibration of the sensor/measurement tool being out of wack. Yes, the
>> test you propose would answer that one way or another.

>With default configuration for lm_sensors, sixty seconds of "stress"
>cause a rise from MB 40, CPU 30 to MB 43, CPU 39. After playing around
>with sensors.conf, most plausible result so far was rise from CPU 43, MB
>29 to CPU 45, MB 40.

No, that is not plausible at all. The more rapid rise is always teh CPU.
Unless your idea of stress was to direct a blowtorch at the motherboard. It
is the cpu that is the source of heat, Under stress it is what gets hot.


>> Yeah, but have you actually verified that the temperatures being read
>> are reasonably accurate? Certainly it's a lot easier to stick a
>> thermometer in between the PCI slots than to jump through the hoops
>> the CPU manufacturer specifies to actually measure the die temperature.

>If I get really serious about this, I'll get a digital thermometer with
>remote sensor, and assume that's the most accurate reading of all.

reading of what? Teh cpu is supposed to be an on-die temp, which your
thermometer will not be able to measure unless you grind off the top of the
cpu package.



 |  Next  |  Last
Pages: 1 2 3 4 5
Prev: Problems with badly burnt DVDs
Next: Salts and farming