Linuxguru: Yes - thanks for calling me on it.
In the same physical package (an IO with the cover on), my K6-III at 200 reached equilibrium at a higher temperature, both CPU and Q16-wise, than the WinChip 200.
When I saw this, I took those "temperature" readings at face value and jumped to the "obvious" conclusion - namely that the K6-III box is generating more heat than the WinChip box, and that it must therefore be consuming more power. (Less power dissipated at the CPU, more power dissipated through Q16.)
I noticed that the curves of the two sensors differed significantly during the course of each test.
On the WinChip 200, both curves tracked each other to within a degree or two from 20C up to around 45-50C final temperature.
On the K6-3 at 200 MHz, the motherboard (Q16) sensor went to the high-50C range relatively quickly, where the CPU sensor (located on the V1 motherboard beneath the CPU) spent a lot of time in the 30s and 40s; it took a much longer time to catch up.
What I think is happening is that the K6-3 is running plenty cool (i.e. consuming less power - as in volts X amps - than the Winchip), but Q16 is also consuming power, and much of this heat is being dissipated into the motherboard. Eventually that heat flows and hits the CPU sensor, causing it to report a higher temperature than if it were tracking CPU power consumption only.
If a sensor half an inch away from Q16 reads 60C, it probably wouldn't be a stretch to see the sensor on the motherboard in the "square" part of the CPU socket (only another inch or two) hitting numbers in the 40C range even *without* a CPU to generate heat.
Unfortunately, that means I can't rely on either "temperature number" to come to a meaningful conclusion between processors.
The way to settle this debate (and actually, this'd be a fun project) is to hook an ammeter in series with the power supply on a Winchip-200 and a K6-3-200. There are no moving parts, so it all gets turned into heat (well, minus the photons in the backlight!) inside the case.
Based on the fact that the K6-3 *chip* heats up more slowly than the WinChip, I think we both agree that the K6-3 *chip* draws less power than the WinChip.
Where we differ is the notion that you believe K6-3-200+Q16 consumes less total power than WinChip200+Q16. If I rely on my temperature numbers to reflect what's going on inside the machine, I think the K6-3+Q16 *is* consuming more power (3.3V x amps) than the Winchip, but I've just established that I can't rely on those temperature numbers to tell me anything meaningful about total heat dissipated inside the case.
So you've convinced me - there are enough confounding factors that I no longer have an opinion either way in terms of overall power dissipation. I'll guess that the K6-3-200 and WinChip 200 are probably comparable, but I won't know which one's gonna come out on top until I actually break out the ammeter and see who draws more current outa the wall.
(Now I just gotta dig out the parts to put the test clips in series with the 19V output from the wall wart without cutting my power cable up... aaaaaah, this IO hacking never ends!)