Petz,
Power dissipation between 0.25um and 0.35um:
By now everyone knows:
P = C * V^2 * f, where
P = power dissipation,
V = applied voltage to the core,
F = core operating frequency, and
C = geometry constant.
In 0.35um K6, V = 3.2V, f = 233, and P = 28.3W (when you design anything to meet a spec, you always have to use the worst case, and I hope you know this), these hints a C = 1.19E-2 with an appropriate unit of Watt/Volt^2/MHz if you want to get down to more details.
In 0.25um,
C = 1.19E-2 * (0.25 / 0.35)^2 = 6.07E-3,
V = 3.2 * (0.25 / 0.35) = 2.3,
And say f = 300Mhz,
We have P = 6.07E-3 * 2.3^2 * 300 = 9.6 watts worst case for K6 with 0.25um process.
Now, what is the theoretical limit of K6 at 0.25um? Take the limit at 0.35um of 233MHz:
233 * (0.35 / 0.25)^2 = 466 MHz.
At f = 466MHz, the power dissipation (again worst case) becomes 15 watts.
Yes, it looks like the Socket 7 is very capable of dissipating this meager 15 watts, but the die size is reduced to 68 / 162 = 0.42, that means the junction-to-case thermal constant is increased by 162 / 68 = 2.4! Now, can you keep the junction temperature at a satisfactory level? Slot I has lower case-to-ambient thermal constant, and that extra bit of advantage helps.
John. |
