Also, something to think about is "bang for the buck".
A while back, I benchmarked the K6-III at various clock rates vs. the WinChip C6 at 200 MHz:
A V1 (WinChip 200) can easily double its speed (K6-III-200) in real-world applications for $30 and some soldering.
While a K6-III is faster than a Rise/P200MMX/WinChip2, the speed improvement isn't quite as extreme as it is vs. the V1's WinChip 200.
If you've got a V1, I'd say the K6-III hack is very much worth the trouble. (The soldering required to change bus speed and/or multiplier is much easier than the soldering required to get the chip up and running in the first place!)
If you've got a Rise or WinChip2, it's still a Good Thing to upgrade the CPU, but it's the night-and-day improvement over the V1.
Turbo3: Very interesting memory bandwidth numbers you've got for the Rise - they're about a third of the WinChip 200 (!) (Disclaimer - my WinChip 200 results were made with 32M onboard, not 128M. That shouldn't affect anything, but who knows?)
ckbone: Well-said. This isn't a mod for folks who've never picked up a soldering iron. (If you've never picked up an iron, get a scrap circuit board from a surplus store and practice on it first!).
While I'm very happy with my K6-III (over V1 C6-200) performance boost, I only did it because I (a) felt comfortable hacking SMT, and (b) was willing to risk my IO in order to have the pleasure of the hack.
What it comes down to is "why do you need the speed"? If it's for games, consider buying a used active-matrix laptop. (Unless you have parts, the cost of a hard drive, power supply, and what-not to run games on an IO will be about the same price as a used laptop, and the laptop will have a better screen). About the only "good" reason I can see for adding extra CPU power to the IO is for MP3 playback or mobile mapping software for GPS.