And to think that just today we were learning about saturation of semiconductor electrons with increasing Vcore :-D

That's what I'd say is happening anyways. They're trying to push the electrons through each gate faster with the move to 2.6GHz, and on this design at least, and with the current transistor technology they're using, they have run into a saturation problem. Saturation causes more collisions with other electrons and atoms, and so more heat. In addition, the FX-55 is doing more work, but I'd say that the increase is more than linear (as increasing clockspeed generally is) :-P

This theory kinda depends on the FX-55 drawing more voltage to power the movement of the electrons when the chip reaches saturation points...

I looked at the TR article, but couldn't find anything, I suspect mainly because the chips were run at 1.6V both times, which could mask the effect. It isn't stated I don't think anywhere what voltage the FX-55 takes either. I looked on the anandtech article - couldn't find anything. Hell, even the tech specs of the FX series hasn't turned up anything yet :-D

However, this review seems to poo-poo my theory:
http://www.sharkyextreme.com/hardware/cpu/article.php/3424131

"Both processors use a 1.5V nominal voltage, but differ in terms of thermal and current requirements. The Athlon 64 FX-55 has a maximum thermal power of 104 W and a maximum processor current of 80A, while the Athlon 64 4000+ has slightly lower 89W and 57.4A specifications."

Either way, I'd say it's drawing more power to get the electrons to go faster :-P
-Mole

damage overclocked the fx-55 to 2.8 with ease, so there may be headroom for one more bump yet . . .

I can think of two reasons why the 90 nano parts might have a lower TCase spec.

- With the smaller die, the thermal resistance of the interface between the die and the underside of the heatspreader is higher. So the temp of the headspreader itself (which is what I assume they mean when they spec TCase) must be kept lower to maintain the same die temperature.

- Aren't circuits made on smaller process geometry more succeptible to detrimental effects from heat? If so, then you would want to keep the chip cooler to prevent problems.

This is expected. Early yields aren't so great, so the 2.6 chips need more Vcore and therefore better cooling. AMD is doing it's own overclocking.

Apart from interesting details on TDP, the Processor Power and Thermal Data Sheet also lists maximum case temperatures (TCASE Max).

The way you have worded it here many people may think you are refering to their ATX/SFF/ect case temps, Tcase, of course, refers to the processor case temp ie the temp of the heatspreader.

that is quite strange... the only thing I can think of is that the 90nm are running cooler but are for whatever reason more difficult to cool than the 130nm parts (i heard due to less surface area)... but im certainly not qualified to comment on this

either way... certainly a lot more positive than the p4 90nm flame throwers, let's hope it stays that way...