At least one of the two "defective" calcs is very carefully handled (the same guy has a TI calc that he is also very careful with). I don't think this is related to the health of the main board because if it was, then IMO the calculator would show problems at other CPU speeds too, or else, it wouldn't stop working immediately after a certain CPU clock is selected.

As AHelper and/or Ashbad were saying earlier, though, part or indeed all of the problems may be the peripherals, particularly the RAM and what sort of access speeds it can stably handle. If they cut corners or used different RAM in some of the models or versions, then it's possible that the RAM (or other core peripherals) simply can't handle the faster speed.

I think I found out why some Prizms won't overclock to 94.3 MHz. On a friend's calculator that has the problem, I managed to overclock to that speed without a immediate system error, hang or reboot. I got to the Menu, and it still worked, except that there were no add-ins (same effect as if you had unmounted the user partition in the Test Mode).

When the calculator tried to access the storage memory again, it just shut down and reboot compulsorily. My conclusion: it is the flash chip of some calculators that can't handle such a high CPU speed.

That is an excellent conclusion, and should allow us to design programs with sections at different speeds that run safely on all calculators. Can we test a program that runs a section entirely in RAM while overclocked and returns to a "safe" speed before going back to executing in Flash, and see if it works on the known-bad hardware?

3-The OS is playing with the rest of the CPG which we completely ignore.