One thing you're guaranteed in a web-review of a CPU, a fansink, a case - any widget in the warm - is that it'll come with a table of
temperature measurements. Sadly, with very rare exceptions these are about as much use as if the reviewer had sidled up to their computer & tried to stick a thermometer up its . . . . .
. . . . ahem.
Half the problem is that we don't all live in Anchorage, Alaska or Panama City - so all "absolute" temperature
measurements need allowances/corrections/guesstimates for the effects of the local environment; the other half is that the equipment & techniques used to obtain such measurements are - in the main - at the Astrology end of scientific method.
Let's put it another way: Intel - who started all this by putting a thermal diode in their recent processors - allow around 5°C
variation, PLUS some variable "motherboard correction factor," in their published guidelines for calculating PIII efficiencies. So, frankly, all these absolute measurements you see on the web taken from uncalibrated mobo sensors, are . . . well . . . let's call 'em well-intentioned
Since we don't all have - never will have - certified calibrated thermocouples - what's needed to share meaningful information
is a portable standard which any of us, anywhere, can use quickly to apply web-review-data to our real-world system - & vice versa, for those many folk who want to get some facts out there.
Fortunately, such a standard measure of efficiency exists, & is easy & cheap to measure or
apply: it's called "°C/W"
Two's better than one:
°C/W is a portable & reliable measurement because it's relative - you don't need expensive accurate thermometers, just a
pair you know agree with each other: the idea as applied here is that you use a cheap easily-available "Inside-Outside" digital thermometer hacked to provide a matched pair of remote bead-thermistors with independent digital readouts. Using these, you compare the heat going in with the
heat going out. What you measure is the case temperature of the CPU & the temperature of the fluid [air, mostly] you're using to cool it, then perform a simple calculation relating these to the amount of power going into the CPU & applying - if you like - a few correction factors.
The number or range you come up with has real meaning wherever you live & whatever your system; you can easily calculate
back from a published °C/W number to see how well the widget in question will work in your own system. The manufacturers of stuff like fansinks usually have a °C/W tucked away quietly somewhere; & you can have hours of fun seeing: a) if that 50CFM fan you've put on their innocent product is
worth the fashion-challenge of living in earmuffs, & b) whether their product information belongs in the fiction shelves.
Any digital inside-outside thermometer with similar bead-thermistors acting as the two sensors can be hacked to give you a
suitable °C/W-measuring device. I very strongly recommend going to www.benchtest.com right now & reading their clear & thorough how-to on hacking a Radio-Shack 63-1024 "Inside-Outside" - this very cheap
device works well for our purposes & is so simple to hack that even a ten-thumbed anti-solderer like me can convert one in about half-an-hour.