lmsensors on GA8GE667Pro

I have no experience with your particular board; but there are a few obvious points and some general ideas that might help a little.

First, this sort of data is valid only as an indicator, not as a specific measurement. What that means is that you don't really care how accurate the voltage reading is, and if your 5 V line says it is 5.00 volts when in fact it is 5.15 volts, it make *no* difference!

The main reason for that is because you don't care if it is 5.0 or 5.2 volts, but you do care if it changes from 4.8 to 5.2 volts! All of those are within specs, but a change from 4.8 to 5.2 is something you might want to investigate.

You have no "compute" statement for in3 or in8, which probably need one. Apparently in0, in1 and in2 do not.

This one appears to be okay, given that the +12 volt line is saying it is 12.22 volts.

This is the -12v line, and *clearly* is not correct, given the -4.17 reading.

As above, this can't be right, given the -5 volt line is saying it is -.68 volts

This one appears to be correct.

I would simply adjust the compute statements to get readings that are very close to what each voltage is supposed to be. (If the +5 volt line reads +5.00, you will *know* if it changes. If it reads +4.88, you might not notice a change to +5.23. Hence monitoring the system is actually better done with "inaccurate" readings!)

Since in1 is monitoring a +1.5 voltage line, it would appear that the above max and min values should be 1.4 and 1.6, eh?

I would think that 5% tolerance for the plus and minus 12 volt lines is much too strict. 10% is more reasonable.

That last one should be "5 * 1.05".

everything else snipped

My experience has been that the expression parser in lmsensors has bugs. The compute statements don't necessarily do what they should, and and tripping a bug in one statement may make everything after that incorrect. Hence, try to keep the expressions simple.

That should probably be changed to

When making changes to the compute lines, there are various ways to change a value. If the raw data is 2.000 and a value is added to it, then a change from 2.000 to 2.100 will cause a 0.1 volt change in the voltage reading. If there is a multiplier, the sensitivity is increased and a raw data change of 0.1 will result in a voltage change that is 0.1 times the multiplier (e.g., 0.3 if the multiplier is 3).

The -12 volt value is computed with this equation:

That will provide a reading of -12.00 volts if the raw value from the sensor is 2.000. However, it's now reading -4.17, which means the raw value from the sensor is 3.023.

If you change offset value of 27.36 to 35.37, the reading will be -12.00. It also means that a change in the raw value, say from 3.023 to 3.123 or to 2.923 would cause the voltage indication to change from -12.00 to -11.42 and -12.95 respectively.

RAW VALUE VOLTAGE READING

2.923 -12.95 3.023 -12.00 3.123 -11.42

If only the multiplier is changed, from 7.67 to 5.08, that would also result in a -12.00 voltage indication when the raw value is 3.023. However, for 3.123 the voltage indication would be -11.49, an for 2.923 would be -12.51.

RAW VALUE VOLTAGE READING

2.923 -12.51 3.023 -12.00 3.123 -11.49

Which one is more accurate? Who knows! It makes no difference though, as the question is which one is more useful.

Think about this one:

RAW VALUE VOLTAGE READING

2.923 -13.00 3.023 -12.00 3.123 -11.00

Which is to say, almost twice as sensitive (the same change in the raw value gives almost twice the change in voltage reading) and yet it is still a linear response.

I would try that, and see how the values look over a "normal" period of time. If you get swings that set off alarms, cut it back to the 5.08 multiplier.

The -5V compute statement needs similar analysis.

--