gallier2 (gallier2.delete@this.gmx.de) on April 9, 2017 5:38 am wrote:
> David Hess (davidwhess.delete@this.gmail.com) on April 8, 2017 10:26 am wrote:
>
> > Was there some reason that the 80186 had to be incompatible with PC hardware's
> > peripheral arrangements? That seems like a deliberate design decision.
>
> I don't really know but I think that the odd one is the original IBM-PC that used some
> cheappo 8 bit peripheral chips instead of the intel recommended ones. Also the addresses
> of the peripherals in the I/O address space are more or less arbitrary, IBM and intel
> engineers did not pick the same even if they'd pick the same chip (8259 p.ex).

The 8 bit peripheral ICs were less expensive and less expensive to use which was also a major reason to go with the 8086 which had an 8 bit version in the 8088 while the 68008 was not available yet. Intel had that in mind and deliberately designed the 8086/8088 to be able to use the existing MCS-80 peripheral ICs.

Or were you thinking of the 8086/8088 support chips? IBM might have avoided those if they thought Intel was charging too much.

As far as address decoding, IBM probably went with the cheap options there also with a sparse and incomplete decoding lowering the number of gates, ICs, and levels needed.

> This said, the 80186 was introduced in 1982. Its characteristics were therefore
> defined before IBM's PC (introduced 1981) was established as industry standard.

That makes sense to me. I had not considered the release dates.