AVR-Chat

stevech wrote:
> I think it's commonly accepted that IBM went with the Intel CPU for market
> timing reasons - to hit the Christmas market window.

That isn't commonly accepted on this planet. Perhaps in some alternate 
universe. I have heard the following information from multiple sources, 
some of whom were working at IBM at the time.

The Intel processor was chosen solely on the basis that Intel was 
willing to license AMD as a second source. Motorola wasn't willing to 
let anybody else make their CPU at the time. Two out of the three IBM 
groups designing a "PC" had used the MC68000 in their design, but IBM 
(like the military) was more worried about second source issues than 
they were about performance, so the IBM-Intel design group won out with 
their inferior design.

I don't think IBM even had the concept of a Christmas market window, 
especially considering that they didn't expect to actually sell any of 
these "personal computers" they had created. They were looking at the 
market that was looking for a little more intelligent 8250 (is that the 
right model number? (I used them but I don't recall the label they had 
on them)) terminal.

> The 8080 and DOS were
> there but Moto was 6 months behind.

Motorola was not six months behind, they just didn't want to license a 
second source supplier. Two of the three groups coming up with the 
design for the first IBM personal computer were using the 68k.

The group I designed a computer with for a final project in university 
used a 68k for the CPU. I think I might have noticed if we had paid $118 
for something that didn't actually exist at the time.

I have no problems with the rest of your mail except...

> If they had not been so, things may have
> gone much differently. IMHO, PC OSes would have become stable and user
> friendly several years sooner had it not struggled with the horrible banked
> memory non-linear addressing and goofy register set architecture of the
> 8080, an 8008 successor.

You have to admit that the 8086 architecture was pretty cool for minix. 
64K+64K split I and D space made for a very reliable Unix-like 
architecture at a time when memory protection didn't really exist in 
other microprocessor systems. Of course, 64K doesn't leave much room for 
very interesting problem sets, but way back then, 64K was still pretty 
big for many computing problems. (We are discussing this on an AVR forum 
after all.) As I recall, cc was a cheat which couldn't actually be 
compiled on a minix box, so it was only ever supplied as a binary.

On the other hand, I first ran minix on an Atari 520 ST, so I started 
without memory protection...

At roughly the same time, Sun was making workstations with a pair of 
MC68000 processors, one running a clock cycle behind the other. Whenever 
there was a page fault on the one running ahead, the second would be 
halted while the first would fix the fault. Then the second would become 
the primary processor until the next page fault, whereupon they would 
flip again.

Then Motorola released the 68010 that fixed the page fault issue.

Then Sun decided to design their own processors.

> But, VisiCalc, which I say "MADE" PCs happen, ran OK none the less, and ran
> on Apples too.

John (who is feeling old and crotchety today, realizing that he has 
lived through more than two thirds of digital computer history and 
worked through more than a third of it (and being paid to learn a new 
language right now, so it's not all bad))