While I'll admit that my assessment of the block may have been flawed, at this point I'm more interested in a determination of what causes the difference.
On the heat die side, 14 by 14 was used (196 mm2), with a 80W (applied) load. (Also a 100W load on a 32 by 32mm)
On the CPU side, the core area is 115 mm2, with max 104 W
Temp measurements are calculated as:
dT ºC (differential temperature) = IHS ºC - Ambient Air ºC
For the sake of argument, I'll add a 1mm strip around the core area, to compensate for the IHS, making it roughly 13 by 13 mm.
Obviously the difference is in the heatload.
not the area of the heat die used.
As far as I can figure, the selection of 80 Watts was designed to replicate:
a Prescott 2M 90nm, which has 135 mm2 core area and 84 to 115W heatload
or
a Sledgehammer 130nm, 193 mm2 / 84.7 W
(not sure which)
What's left to discuss then is; what heatload should be used, if one is using a heat die, since it's obviously going to make a difference? (as if we didn't know)
I'd like to see Lee (Robotech) run a test series with the 14 by 14 die, at 100W (I suspect that the PSU may be limiting here
). I know its a lot of work though.
It also does not make sense to me that the difference (between Storm and Apogee) would be greater with (essentially) the same core area, but with a smaller heatload. Is the temp measurement on the side of the IHS a reliable measure? (it's very sensitive to airflow) Is the core temperature data of any value? (it seems to be relevant). Why did over-volting the CPU (essentially) not make any difference (between blocks)? What would be the estimated power load while over-volted? (125 W?)
I'm left with more questions than answers.
