At Hackaday Belgrade 2016, Mike Szczys spoke about choosing between 8-bit and 32-bit processors. He made some good points about the costs of optimizing to the cheap processor before you have a good handle on your project. We’ve done that, been advised about that, and advised about that.
Mike touched on how small processors are easier to learn, but once you get the hang of this gig, the power tools of the big processors give you the leeway to put together more complex projects faster without killing yourself.
It’s like the Arduino, which is easy to learn, but you outgrow it quickly. As you get closer to the limits of your processor, effort goes to infinity (and beyond).
Just the other day, one of us was trying to coerce our SwornEnemy into using a processor too small for his needs, just to drive him mad (actually not). Since he was just embarking on his task, having a too big processor with too much RAM/ROM/peripherals is the way to go. Since the the board will require another spin later anyway, it’s easier to go for a cheaper processor than beg for a more expensive one.
You can find more about the the talk on Hackaday.