I don't quite trust lm-sensors and am further a little frustrated that this is the only temperature monitoring system that exists for GNU/Linux, a platform supposedly developed by geeks and tinkerers. Don't geeks and tinkerers care quite a bit about monitoring their CPU temperatures in a trustworthy manner?
When I was on Ubuntu I remember lm-sensors actually showing lower temperature values than on Debian, which is actually running less processes at a time and should therefore heat the CPU less as well. It doesn't make sense.
But the case against lm-sensors doesn't stop there. I did a quick Windows XP install on another partition and installed some popular and often quoted temperature programs for windows, coretemp, realtemp, speedfan and Intel's own thermal analysis tool. These are the programs that CPU related hardware reviewers always quote when measuring their temperatures so what these programs show seems pretty trustworthy, especially when they all or most of them actually agree on the temperatures they are showing (yet aren't all using the same system AFAIK).
All of these programs are showing temperatures by almost 10 degrees lower than lm-sensors in Debian does, which makes a compelling case in their favor and makes me think that I just can't count on what lm-sensors readings and would be closer to the truth if I substracted up to 10 degrees from what it's showing.
I wonder if anyone else had similar experiences? It's a little shameful that GNU/Linux is so weak on this point.