Great, thanks for all of the info. I'm pretty sure I can knock off about 5 degrees by correctly applying the grease and waiting a few days, but since I have new stuff coming in a day or two I'll just wait. For anyone that's interested I'll repost my readings.

in6 is maxxed out -- all 255 bits set -- so is meaningless. Each bit corresponds to 0.016 V. The chip manufacturer recommends that unused sensors are earthed but this seems not to have been done by the motherboard designers.

fan3 is present on the it8720 monitoring chip which is what lmsensors is getting the data from. lmsensors has no way to know whether the fan is connected except for the sensors3.conf file in which it shoud be told to ignore fan3.

At least one of the voltages will be nominal 12 V and needs scaling in sensors3.conf to give a meaningful reading.

Be careful using tower heat sinks because motherboards are designed to be cooled by the CPU retail heat sink. This means if you are using a tower heat sink the air is that exiting the tower heat sink is not directed to the motherboard, so your motherboard will eventually over heat.

I checked the thermal limits and it saids 70 degrees C. This means at this temperature it will damage the processor and it is close enough to damage or worst has been damaged. If anything have to throttle to decrease temperature, the device is already damage and life expectancy is penalized. Thermal throttling is just a band-aid and not a real fix to protect the processor. If thermal throttling feature have to be used, it is best to throw away the processor and buy a new one noting maximum temperature.

When I said that 50 degrees C is OK, in engineering sense the processor is very, very close to damage because there is not enough head room if the processor is working at 100%. Really there is no average temperature that any processor should be at. It best to work the processor at 100% and make sure it does not hitting 60 degrees C. Then you know your heat sink is doing its job.

All thermal sensors are inaccurate. Thermal sensors just gives an average rating of the surrounding components. The temperature can be either high or low, so this is why they are inaccurate. Also thermal sensors have to be everywhere (top, bottom, and in between) to gather the information. Unfortunately, you can not get a reading of the hottest part which is in the middle.

No we do not need to know the average life expectancy of this processor since we are not designing a satellite to have to last 50 to 100 years. Your question catkin is completely irrelevant. I already stated the answer.

Good info, thanks electro

The N and S bridge have pretty sizable heat sinks and I have a 120mm fan blowing through the case, do you really think this will be a problem with a vertically oriented cpu fan?

Ok, I'll have to look a lot deeper into lm_sensors. This is my first foray with it.

If "vertically" is not toward the motherboard then yes, it's risky.

Generally motherboard manufacturers provide BIOS that shows everything that the hardware measures so a good place to start is in the BIOS, both to see what is measured and associated values. Some BIOSes use the CPU in an infinite loop so the CPU temperature can be misleading but probably not on your motherboard.

The next step is to look for a sensors3.conf file for your motherboard. In the (sadly likley ) event that you don't find one a good starting point is a sensors3.conf file for a similar motherboard. Motherboard designers tend to re-use existing designs, evolving them as new chips come along.

Good places to look for sensors3.conf files: http://www.lm-sensors.org/wiki/Configurations, http://lm-sensors-db.berlios.de/ http://www.lathspell.de/linux/lm_sensors/ and http://khali.linux-fr.org/devel/lm-sensors/.

If those steps don't provide you with credible results and you need to dig deeper ...

The analog voltages are converted to 8-bit digital values with a default of 0.016 V per bit giving a range 0 to 4.08. When greater ranges are required, as for 5 V and 12 V values, "potential divider" circuits are used at the inputs. The best explanation of these, in an lmsensors context, is on the gkrellm man page, in the "Voltage Sensor Corrections" section.

When a potential divider circuit is used an equivalent compute line is required in sensors3.conf. Ideally this is based on the values of the resistors used in the potential divider circuit. These values are seldom published. Happily designers usually follow the monitoring chip makers' recommendations when choosing values. Compute lines for chip makers' recommended resistor values can be tried and then the voltages displayed by the sensors program checked against voltages in the BIOS screens or measured by voltmeter.

New heatsink and as5, just after install:

I'm not sure what to make of the 3 different readings. my BIOS only shows 2 temp readings. Although supposedly that thermal diode (temp2) is much more reliable than the thermistor (temp3). In any event, my system is running cooler, and quieter. I'm happy. After the break in period it should be even better.