Hallo Tobias, On Tue, 13 Jan 2009 12:20:46 +0100, Tobias Preclik wrote: > thank you for your detailed explanation. The problem seemed to be fan4 > indeed. I updated lm-sensors to 3.0.2 though it is still marked as > unstable (and hard-masked because barely any applications can deal with > it) in Gentoo. FWIW, lm-sensors 3.0.2 (and later) are actually very stable. The only problem is - as you noticed already - the lack of application support. But I expect this to be solved by the end of 2009. > But now sensord starts without complaints and the fan4 is > automatically non-existent. But still the values reported by lm-sensors > seem bogus. Especially +5V and temperatures values: > > w83627ehf-isa-0290 > Adapter: ISA adapter > VCore: +1.59 V (min = +0.00 V, max = +1.85 V) > +12V: +12.04 V (min = +10.82 V, max = +13.20 V) > AVCC: +3.38 V (min = +3.14 V, max = +3.47 V) > 3VCC: +3.36 V (min = +3.14 V, max = +3.47 V) > +5V: +4.30 V (min = +4.76 V, max = +5.25 V) ALARM > VSB: +3.38 V (min = +3.14 V, max = +3.47 V) > VBAT: +3.28 V (min = +3.14 V, max = +3.47 V) > CPU Fan: 2481 RPM (min = 1704 RPM, div = 8) > Sys Temp: +27.0?C (high = +0.0?C, hyst = +0.0?C) ALARM sensor = thermistor > CPU Temp: +9.0?C (high = +80.0?C, hyst = +75.0?C) sensor = diode > AUX Temp: +77.5?C (high = +80.0?C, hyst = +75.0?C) ALARM sensor = thermistor > > How should I go about tweaking the temperature and voltage computations? > Should I measure the 5V lines or just fiddle about the resistor values > and pretend the raw voltage value is 5V? Voltage configuration is motherboard-specific and can be a little difficult to figure out. A general advice is to check the values reported in the BIOS and see how sensors.conf can be tweaked to come up to approximately the same result. I agree that your +5V is pretty low with the default scaling factor, and that's rather unlikely to be true. See what your BIOS says. The other Asus boards using this chip use: compute in5 @*(1+(22/10)), @/(1+(22/10)) One trick that worked many times for me is to wait a bit in the BIOS until the reported voltage changes. If your PSU isn't perfect, this should happen. Then you have 2 different values for +5V, and the difference between them is equal to the ADC's LSB value * the scaling factor. The W83627EHF's has an ADC LSB value of 8 mV, so you can deduce the scaling factor. > What about the temperature values? The CPU temperature seems to be too > low (though the room is admittedly cooled). Unfortunately the > /proc/acpi/thermal_zone directory remains empty on this machine and I > cannot easily doublecheck the temperature. Maybe I should check the > values reported in the system BIOS... > > What about the sys and aux temperatures? I seriously doubt that anything > is 77.5?C hot in that machine while idling. I guess the sys temp is a > dead input because the rrdcgi charts show a perfectly constant value of > 27.0?C. Is that reasonable? Which temperature sensors do "sys temp" and > "aux temp" correspond to anyway? First of all: don't trust the labels, they are again motherboard-dependent and the default might not match your system. What CPU do you have in the machine? Maybe you can get either the k8temp or the coretemp driver to work. I don't have much to propose for this part, sorry. But it would be interesting to know what temperature labels and values your BIOS displays. -- Jean Delvare
