> < It's certainly *possible* to add intelligent I/O channels to an x86 mach
> The hottest machine I've played with was a hacked xerox XP12 laser printer
> controller. It was a 8mhz 8086 and a 8089(IOP). It could beat any 8086
> system cold (using s100 and multibus systems as standard). Later I would
> see a 8086 multibus system with two 8089s running CPM-86 and it was far
> faster than the then new AT.
Thats it - I still go mad when people are ranting about the x86
architecture (kind of popular) but they never tried to understand
the features, or they never seen the whole family. In fact, Intel
designed the 86 as part of an chipset in a very mainframe like
way - A general purpose CPU: 8086, a specialized floating point
add on: 8087 and a versitale IOC: 8089. A structure as found on
most mainframe at this time (Remember, also IBM offered their
customers to get specialized add ons to improve floating point
performance on the /370). Since IBM used only the CPU and FPU,
the actual PC design is exact like a Mainframe without an IOC:
Just painfull slow.
Gruss
H.
P.S.: The hate-segmentation-rantig against the x86 also drives
me mad - the segmentation sceme used is a very good compromise
between usability and performance. Loosing up to 15 Bytes
per segment isn't realy a drawback compared to granularities
of 4 or 8K today ...
--
Ich denke, also bin ich, also gut
HRK
Received on Tue Oct 20 1998 - 07:01:17 BST