The calculation of a C/W for a running system is really tricky. The better your wb the higher percentage of the CPU's heat that goes into the water instead of the mobo's traces. The design of the mobo (layout of VRMs and other hot bits; cooling of the northbridge, etc etc) also will affect the heat load. It's not easy to do at all.
Now presumably if you use the same "W" every time then you're just adding a standard error into the results. The concern of course is that the whole purpose of a "C/W" is to produce results that are usable by other people. They aren't often qualified as "C/W*"
*for an IDENTICAL system in every respect including bios rev, psu, ram, mobo, cpu, video card, etc etc
I find the use of C/Ws for testing on real systems to be somewhat disingenuous because you're putting another layer of math between the user and the actual results. It isn't the case in this review because the raw numbers are still reported, but it often happens.
|