A little bit of history since I was there at the time (my office was
across the hall from Dr. Dave Bradley...famous for inventing Ctl-Alt-
Del).
The IBM PC was primarily a skunk works project. As such it had to be
done with as many parts that could be re-used/modified from existing
designs within IBM. The original PC team (17 people if I recall) came
from System/23. The IBM PC bus is the same bus as the
System/23
(electrically and general board outline). To keep from using
(cheaper) PC cards in the System/23 the connectors were rotated 180*
so that they physically wouldn't fit in the System/23 (or visa-
versa). Many of the original IBM PC boards were System/23 boards re-
layed out.
The System/23 is historically significant because if IBM had never
built it, they *never* would have done the IBM PC. It is notable not
only for the above but because it was the first IBM product that used
a non-IBM microprocessor (8085). The Displaywriter was started at
about the same time as the IBM PC but used a much more expensive 8086
vs 8088 that was going into the IBM PC.
I remember a meeting with Marketing where they were apologizing
because they had created an estimate for the total production life (at
the time product life times were ~5 years) of the IBM PC (which is how
the price of a product was set in IBM at the time). In order to get
the price breaks that were needed, large volumes were required.
Marketing came back with the largest number they had ever issued for a
computer product ... 5000 ... over 5 years. Yea, right. I think we
shipped that many in the first month of full production. :-)
On Apr 24, 2008, at 6:25 PM, Eric Smith wrote:
Jules wrote:
Why they didn't do a lot of things when it
came to the PC's design is
beyond me (even down to the choice of CPU - wasn't the m68k generally
available by the 1981 launch of the 5150?)
Sure, but the MC68000 was *much* more expensive than an 8088 at that
time, and required fewer support chips, because Intel had a bunch
of peripheral chips that worked almost gluelessly on the 8088.
I'm not sure whether the MC68008 was available then, but with the
MC68000 the minimum memory configuration would have been 32KB rather
than 16KB, and memory would have had to be expanded in increments of
18 chips.
Most significantly, though was that IBM had already done 8086 designs,
and the PC design more or less leveraged what they already had. I'm
sure they would have liked to have put in a better processor, but
since the PC was just supposed to be a quick-and-dirty hack to show
that they could build a microcomputer, I don't think they were all
that concerned about whether the decisions they made at the time
would have unfortunate consequences later on (e.g., the stupid
edge-triggered interrupts, where anyone with the slightest bit of
experience would have known to use negative level triggered
interrupts).
They clearly didn't expect that the product they were designing for
release in 1981 was going to set the standard for the most popular
computer architecture on the planet for not just the next 27 years
but for the forseeable future beyond that.
Fortunately over time we have shed a lot of the legacy baggage,
including the aforementioned interrupt stupidity. The x86 processor
has even become more friendly to programmers, in the form of the
32-bit flat mode addressing of the 386 and later, and even more so
in "long mode" on the x86-64.
Eric