Well then, I ask you the question how bad is the thermal conductivity of the chip?
Since the chip is physically mostly silicon with a little metal and other bits, and the thermal
conductivity of Si is ~149 W·m−1·K−1, it will probably be around that.
However, the chip is the heat
source and only conducts its own heat outwards, and I
have yet to see how this information helps you overclock.
How many watts are you pulling at 4.8GHz?
Depends on your voltages.
What temperature do you get at the base of the heatsink?
Depends on your voltages.
It should be pretty easy to estimate the power draw of the chip at 5.5GHz if you know the voltages needed relative to 4.8GHz.
No. For one thing, each chip is different, requiring different voltages for different clocks.
Secondly, the correlation between required voltages and reachable clock speeds is
non-linear, thus making predictions very tricky.
EDIT:
To clarify: You might get 1.5x the clock speeds for 1.3x higher voltage, but to get
1.6x clock speed you might need 1.7x higher voltage. And that might just be for
one single chip, on another chip you might get 1.5x/1.4x and 1.6x/1.8x. I know
these numbers are not realistic, but they should get the point across.
/EDIT
Then you can estimate the TJ from the thermal conductivity relative to 4.8GHz and lets say 4.0GHz.
It's an interesting idea, but doing this with any sort of reliability would require you to
properly investigate the above mentioned relationship and several other factors in high
volume sample tests on a large scale with great precision, then doing some serious
academic statistical analysis on that sample.
Basically, you would need to investigate all those ugly, non-linear relationships and
how they behave over the clock, power and voltage spectrum.
EDIT:
And even that would only enable you to predict anything with limited reliability.
However, provided your sample was large enough and you did your statistical
analysis right, you would be able to assess how reliable/unrealiable your
prediction would be.
Also: The thermal conductivity of the chip would not change over the clock spectrum.
Thermal conductivity has
some dependence on temperature, however in the
temperature spectrum we're talking about here that's probably negligible.
I fail to see how changing clock speed in and of itself would influence the chip's
thermal properties. And it also won't be changed by changing the voltages, just
as a side note. The thermal conductivity of silicon is relatively constant in the
environment we're talking about here.
/EDIT
Anyway, so you'd say I can get to the thermal limits of the chip without frying the power delivery or vreg?
You will most likely fry your CPU long before you will ever fry the M/B power delivery
components.
On another note: Did you put the OC Noob thing in your sig or was that a "gift"?
EDIT:
The
only way I know of (and if there are others, somebody please
enlighten me, I'm a curious person) to get an IB chip to 5.5 GHz is with some serious
sub-zero cooling. Dry ice, liquid nitrogen and that stuff. No amount of conjecture,
guesstimates or clever thinking will get you around that afaik. It's just basic physics
(well, thermodynamics, really).
/EDIT
