>I am just curious. >I have a dual AMD motherboard, gigabyte. >It has a winbond chip, w83627hf > >Why do the temperatures always seem skewed - they seem low...? >Why wouldn't the winbond chip just return the correct temperatures? >(not something that needs to have a calculation after it is returned >to userspace) > >How would I go about verifing the temperatures are correct? Maybe my >numbers are correct... > >Sample output: >joe at jedi /usr/src/linux-2.6.9-rc4/drivers/i2c/chips $ sensors >w83627hf-isa-0290 >Adapter: ISA adapter >VCore 1: +1.52 V (min = +0.00 V, max = +0.00 V) >VCore 2: +1.54 V (min = +0.00 V, max = +0.00 V) >+3.3V: +3.25 V (min = +3.14 V, max = +3.47 V) >+5V: +4.84 V (min = +4.76 V, max = +5.24 V) >+12V: +11.67 V (min = +10.82 V, max = +13.19 V) >-12V: -12.61 V (min = -13.18 V, max = -10.80 V) >-5V: -5.25 V (min = -5.25 V, max = -4.75 V) >V5SB: +5.24 V (min = +4.76 V, max = +5.24 V) >VBat: +3.01 V (min = +2.40 V, max = +3.60 V) >fan1: 0 RPM (min = 0 RPM, div = 2) >fan2: 0 RPM (min = 0 RPM, div = 2) >fan3: 0 RPM (min = 0 RPM, div = 2) >M/B: +29 C (high = +45 C, hyst = +42 C) sensor = thermistor >CPU1: +32.0 C (high = +80 C, hyst = +77 C) sensor = thermistor >CPU2: +42.5 C (high = +80 C, hyst = +77 C) sensor = thermistor >ERROR: Can't get VID data! >alarms: >beep_enable: > Sound alarm disabled Easy. The reported temperatures are those of the CPU socket, below the CPU. Not the internal CPU temperature which would be way higher and change much faster. >if i use a calculated value like from another chip in sensors.conf i get >these: >M/B: +52 C (high = +45 C, hyst = +43 C) sensor = >thermistor ALARM >CPU1: +53.3 C (high = +80 C, hyst = +77 C) sensor = thermistor >CPU2: +56.8 C (high = +80 C, hyst = +77 C) sensor = thermistor These computations are weird ways to approximate the CPU temperature from the socket temperature. They do not correspond to a real temperature. Some motherboard makers (Asus...) do such computations in the BIOS setup screen too (and after that, user complain because lm_sensors yields different values). I don't think it's clever. Internal CPU temperature changes very fast (like +10 degree C in a second), while the socket temperature varies very slowly. Also, there is no simple correlation between socket and CPU temperature. Depends on the CPU, heatsink, fan and ambient temperature. I prefer to have a real, physical temperature value for the socket than a guessed (and wrong) value for the CPU temperature. >Also, my motherboard manual says I have a w83627hf, but lm90 also >works, but it only displays MB temp and CPU1 temp and from what I >understand, lm90 does not support 3 temperatures... What is your motherboard brand/model? You are right, the LM90 only supports one local and one remote temperature sensor. However, it is possible that the LM90 is multiplexed, either on the SMBus (two LM90 chips) or at the diode level (one LM90 but two thermal sensors). Please provide the output of sensors-detect (after unloading all the monitoring chip drivers). It's kind of fun since we were just discussing the way we are going to handle the SMBus multiplexing on another board (and all boards which need it in general). Jean Delvare
