I reckon that much of the bad press the eMate got for its slow processor came from people who picked one up to try the handwriting recognition, and finding it slow (which it is - real slow, and slow at non-cursive handwriting recognition at that, since it doesn't "do" cursive), put it down and said, "Nah, too slow". I would say, by contrast, that these people were too quick to jump to that conclusion, and that it shows that they never actually used the machine for its intended purpose.
I bought one for my daughter this year for a series of reasons that i need not go into, thinking all the while about how slow it was, and that maybe it would not be up to much, but discovered when I actually picked it up and tried using it that I really liked it - so much that I bought another one for myself. They are, after all, dirt cheap.
I have a lot of experience with Newtons, having owned one since 1998 (a MP2000, and then a 2100). The main reason for the faster processor in the MP2000/2100 was for handwriting recognition; the eMate, however, doesn't really need good handwriting recognition because of the keyboard, and for Appletalk networking and so forth it is not appreciably slower than the high-end 2000 series, as the file sizes are so small. So the main issue was and is handwriting recognition. The fact is, though, that not only do you not need handwriting recognition on an eMate, but you very seldom ever would want to have it, really, as in actual use of an eMate you almost never write words with the stylus. All you miss, then, is the "Gee-whiz" factor of having it reckon with your scrawl, which is scarcely something that speaks to the functionality of the machine. You would draw on an eMate, of course, and tap, and so forth, but for these uses the machine responds to the stylus quickly enough. I find no notable difference in drawing etc. on an eMate and a MP2100.
To offer some perspective, I find, for instance, that using the (2D) graphing calculator on the eMate gives perfectly acceptable results - and presumably that task is reasonably processor-intensive. (We use one of the machines to help with the secondary school mathematics for my daughter.) You get a second's delay or thereabouts in graphing a reasonably complex algebraic expression, and then the screen redraws nicely. My eMates have the memory upgrade, mind you, which also widens the data path from 16 to 32-bit, and things might be different on a stock model.
I should think that serious hacking would be needed for a logic board transplant, as ports would not line up (in fact, they are different on the eMate), the pcmcia cage is different, the keyboard connection most likely does not even exist on the 2000/2100 board, etc. etc. I doubt that it would be do-able, but more than this, I don't see any reason for wanting to do it, beyond the fun of tinkering, as the eMate is a good little machine just like it is.
