Page 1 of 1

.

Posted: Thu 09 Feb 2006, 02:20
by brianez21
.

Posted: Thu 09 Feb 2006, 04:57
by Flash
I don't know if there is a way in Linux to monitor the CPU temp, but the VIA C3 on the EPIA 5000 uses so little power it would probably be pointless to do it.

I replaced a 733 PIII with an 800 MHz VIA C3 to try it out. The C3 chugged along fine with only a small heat sink and no fan. I removed the heat sink for a minute or two, just to see what would happen. The CPU got too hot to hold my thumb on but it didn't quit, and it showed no ill effect after I replaced the heat sink.

Posted: Thu 09 Feb 2006, 07:07
by Sage
Flash is entirely correct. The whole thrust of the NEC/VIA C-series is/was to build low power cpu s, so thermal monitoring is largely irrelevant.
In general, however, it is unwise to rely on SW thermal management unless you have access to a lab. which can calibrate the system. It is necessary to read the full technical manual for the cpu as well as m/b because there are several entirely different internal and external methods used to derive a figure for temperature. It is also essential to distinguish between core and cpu top surface temperature - not always clear from published brief specs. Instead, drill a small hole from one edge into the base of the HSF and push in a thermocouple. There are plenty of cheap plastic DMM s around with a temperature function.

Posted: Thu 09 Feb 2006, 16:05
by brianez21
.

Posted: Thu 09 Feb 2006, 16:30
by Sage
Like I said, you cannot rely on SW monitors. You really have no idea what they are reporting without lab. calibration and detailed manuals data, although they can sometimes be useful for following effects of major HW mods.
If you just want some big numbers to gain a bit of street cred to show your mates, write yourself a little window that reports some amazingly ficticious performances.

Posted: Thu 09 Feb 2006, 17:41
by brianez21
.

Posted: Thu 09 Feb 2006, 21:51
by Guest
Does the motherboard/bios support ACPI? If so, check out this thread.
Wih ACPI properly enabbled, wmpower will allow you to monitor the processor temperature as reported by the machine - with all its llimitations.
As an alternative, many BIOS readouts show the temperature, so you could check that out by rebooting after the machine after it has been running for a while.
Sorry if these are unsuitable suggestions, but I know very little about the VIA EPIA series.

Posted: Thu 09 Feb 2006, 22:05
by brianez21
.

Posted: Fri 10 Feb 2006, 06:50
by Guest
brianez21: you don't seem to be understanding what I said? You are no further forward. Getting some SW working doesn't help. You have no idea what it means!!!!!!!!
Use a thermocouple.

Posted: Fri 10 Feb 2006, 08:47
by Sage
This website is driving me crazy!!
I WAS logged in when I penned the above comment. And, even though I ticked the 'stay logged on' box, I had to do it all again at this attempt!

John! Please!! Be reasonable.

Posted: Thu 23 Feb 2006, 16:13
by miq
The VIA EPIA V5000 (eden) and V8000 series mainboards do not have thermal monitoring functions.

Posted: Wed 01 Mar 2006, 21:02
by brianez21
.

Posted: Thu 02 Mar 2006, 00:13
by Flash
Sage wrote:brianez21: you don't seem to be understanding what I said? You are no further forward. Getting some SW working doesn't help. You have no idea what it means!!!!!!!!
Use a thermocouple.
Sage, I think the temperature is probably inferred from the voltage of a P-N junction diode that is on the CPU die for that purpose. It doesn't need to be calibrated; the forward-biased voltage-versus-temperature of a simple silicon diode is predictable enough that a temperature reading based on it will be accurate enough to be useful. Bardeen, Brattain and Shockley got a Nobel prize in part for proving that.

Posted: Thu 02 Mar 2006, 06:27
by Sage
Sorry, but you are not correct. The thermal response of a p-n junction is widely used in all sorts of sensors. However, everything needs calibration. AMD have been etching an internal diode for a while, but there are issues with calibration and interpretation, as you can see from their datasheets. A voltage is not a temperature! Somewhere along the line you will need a thermocouple and a cold junction. If you want an approximation, stick a cheap thermocouple into the HS and connect it to a cheap DMM to get a precision and reproducibility within about one degree K. Notwithstanding, some folk won't listen to good science however many times it's explained to them!