?

Log in

No account? Create an account
1337ification of phobos, temperatures - Alex Belits
abelits
abelits
1337ification of phobos, temperatures
I have brought a thermometer from work, and measured the temperature of the CPU. Thermometer's sensor was taped to the CPU under the heatsink, so even though this setup is imperfect, it gives the measurements closer to the truth than sensors _under_ the CPU. The outside temperature was 28C, and the case was open. Fan speed, diode and thermistor readings were taken from lm_sensors. BurnK7 (from cpuburn) was running to create the load.

SettingFan speed,
rpm
Temperature,
degrees C
Diode
measurement,
degrees C
Thermistor
measurement,
degrees C
"Low" speed34005646.042.0
"Medium" speed51005142.540.5
"High" speed65004841.539.0


Conclusion:

1. The heatsink is fine, and it's safe to run the fan at the "low" speed, at which it's pretty quiet (at least while the fan's bearings are new).
2. Motherboard sensors' reading may have something to do with actual temperature, however they certainly AREN'T that temperature.
3. Having a lot of air and space really, really, really simplifies the heatsink.
3 comments or Leave a comment
Comments
hollow_hope From: hollow_hope Date: September 21st, 2004 08:39 pm (UTC) (Link)
Are you using lmsensors to get those readings? Remember that the values reported by the sensor chips are not in a "degrees" format at all, and the value -> temperature conversion can vary my manufacturer and what they have hooked up to the sensor IC.

lmsensors tries to make a "best guess" conversion table for the values but it is often wildly inaccurate.

If you use the mobo manufacturer's sensor utility under Windows ( gack ) it will probably yield more accurate results.

(or maybe not, just my thoughts and my experience with lmsensors)
hollow_hope From: hollow_hope Date: September 21st, 2004 09:07 pm (UTC) (Link)
(And somehow I missed that you were, in fact, using lmsensors, so yes I think there's your problem.)
abelits From: abelits Date: September 26th, 2004 04:12 pm (UTC) (Link)
I thought that it may be a part of the problem, but it surprised me that that both sensors show something non-linear -- I thought, at least nonlinearity is compensated in the chipset. Now I have just built a dual-boot box with the same motherboard, and the temperatures still don't look right in Windows utility (just like they didn't look right in the BIOS monitor) -- I should look, just how different is what it shows from what lm_sensors see.
3 comments or Leave a comment