CPU heat output and C/W question

[fvr]

Hi

I'm trying to figure out the C/W value for my home-made CPU waterblock and while banching/stressing the CPU this is what I found:

Configuration #1:
At 2.187GHz 1.85V (13x167MHz)
Tmax=44.5C
T idle=37.0C

Configuration #2:
At 2.2GHz 1.85V (11x200MHz)
Tmax=38.0C
T idle=31.0C

Configuration #3:
At stock 1.83GHz 1.65V (11x167MHz)
Tmax=36.0C
T idle=33.0C


Water T=23C <<<not accurate, I’ll post correct temp tomorrow
CPU: Barton 2500+
MB: Abit NF7-s bios version 19
Pump: L30
Flow rate: haven't tested yet

As you can see the heat output at 200 FSB is significantly less than 167MHz while the CPU speed remains the same. I don't know if this is common knowledge or not, but what I need to know is which configuration of FSB/multiplier should I use to calculate the heat output? The problem is that I need to know how much heat my CPU is producing to calculate C/W. So I use the following formula:

Watts = [(Deff. Watts)x(OC speed in Mhz)/(Default speed in Mhz)]x(OC Vcore/Deff. Vcore)²

I'm not sure who suggested the above formula or how it was derived, but the formula probably assumes multiplier overclocking.

Basically, I need some clarification on the above formula and an explanation for the difference in temperature between the two FSB frequencies.

My second question is about C/Watt value. Since I’m using Prime95 Torture test (small FFTs), I’m pretty sure that I’m not actually using 100% of my CPU. Some CPU power must be reserved for Windows. Since less heat is produced by the CPU, should I correct the answer from the above formula so it can be used in the C/W calculation? For example reduce 100W by 10% to 90W and then divide dT by this value? (where dT=T(cpu)-(Tfluid))


Thanks