There are reasons why one might not want to do that ...

#1 is I'd never pay more than a buck or two for a board with an MCU that doesn't allow separation of the async comm port that the board-resident monitor uses from the serial ports that are on-chip, for reasons already discussed. Sure, I've got 'em, but they were either freebies, 'cause that's what they're worth, or because I got 'em for a buck or two.

#2 is that I'd never buy a board with costly ultra-slow analog features that I'd seldom be able to use (a) because they're too slow for what I need to do, and (b) because I could never sell 'em because they're too costly. Yes, I've got some of those, too, but I've never found a use for 'em.

#3 is I'd never buy a board just to put it on the shelf on the off-chance that I might use that MCU version some day.

#4 is I'd not be inclined to buy a board that didn't have at least 25 square inches of prototype area, properly supplied with both power and ground, and plated-through holes so the pads don't lift if I have to switch components/sockets.

Now, I could go on ...

That, in fact, is why I'm rolling my own. However, I'm still not intersted in any fine-pitch SMT or BGA packages. If I want to use that technology, I'll use a soft-core in an FPGA and be done with it, at least until on-chip ADC's and DAC's convert at >>100 MSamples/sec and cost less than $0.25 per bit. In the meantime, I'll use external converters, so I can point to the converter and say, "See that? That's why this thing costs so much. I'm certainly not going to put a chip with converters into an application that can live without 'em until they're so cheap they don't cost more than the bare CPU, and until it doesn't cost any more to use them than the components without which I can't finish the job.

Now, if they had a real on-chip PLL or two not dedicated to anything else, specifically, ... with an acquisition time of <5 uS, low phase-noise, and a decent tracking range, at, say 20-25 MHz ... that might be interesting.

RE