"Opinion: Critique of the F8 Microprocessor"
By Dave Caulkins
Computer, Vol. 10, No. 8 (August 1977): 83-86


This document was uploaded to scribd.com by "John Strupat."  The original four 
jpg files can be downloaded directly from here:

http://www.scribd.com/jstrupat/d/46923120-Fairchild-3950-Review


Introduction
------------

The Fairchild is a control microprocessor whose architecture is considerably 
different from most machines in the same price and performance class.  Good 
computer architecture is consistent, symmetrical, and coherent; the programmer 
is provided with the maximum amount of information possible after each 
operation and his freedom of action is limited as little as possible. A machine 
with these qualities behaves in the way one expects it to behave; it is free of 
special cases and peculiar quirks. The F8 falls considerably short of meeting 
these goals. It predates most equivalent micros; its designers seem to had 
little contact  with designers doing similar things This background has 
resulted in a machine combining brilliant design concepts with ugly flaws.

Two of the brilliant concepts deserve special mention

1) The F8 CPU (3850) has allocated a scarce resource, package pins, in a novel 
way. Instead of a conventional address bus, there are two 8-bit I/O ports. A 
bidirectional 8-bit data bus and eight control lines provide all required 
communication between the 3850 and associated
Program and data counters (PC and DC respectively) are resident OUTSIDE the 
3850.

2) The 3850 has 64 bytes of resident scratchpad memory. The power of these 
features is indicated by the fact that similar controller-type machines of much 
more recent., vintage (the Intel 8048 comes to mind) provide very similar 
capabilities.

The F8 flaws are covered in detail below. On balance, I think the advantages 
outweigh the flaws, but they turn the F8 from what might have been a really 
nice machine into one which is cost effective but exasperating to use.