Just a month ago, I added one little function to a box I built in '78, which, in this particular case, doesn't use an 805x (it uses an i8080 on an old Multibus-I card) and doesn't require a hardware change at all, but did require the addition of a little code and a cable. It uses one of the two on-board serial ports to notify the control console in the office that one of the machines in the shop is about to run out of stock. That function was not available on the '78 version of the machine, which stamps springs out of a roll of stock, but, by means of a parallel port bit attached to a switch on the machine, adding it was easy enough once the decision was made to do that.

That's an example. It's just a code change, but many code changes have performance impacts that can easily be rectified by means of a CPU speed change, i.e. you can get more done between events X and Y if your CPU runs faster, thereby allowing you to do things you otherwise couldn't do within a given hardware scheme, which, without the performance increase, wouldn't be so easy.

I remember one tape controller that required that the CPU run for somewhat over 8K instructions (about 24kB) in a straight-line sequence with no interrupts or branches outside that sequence. Doubling the CPU speed allowed for loops that collapsed the code length to just under 128 bytes. Sometimes it's worth it to do things in that way.

RE