Path: utzoo!attcan!uunet!husc6!bu-cs!madd
From: madd@bu-cs.BU.EDU (Jim Frost)
Newsgroups: comp.sys.ibm.pc
Subject: Re: GATHER and say NO to MCA!
Message-ID: <23590@bu-cs.BU.EDU>
Date: 2 Jul 88 20:13:59 GMT
References: <42900016@uicsrd.csrd.uiuc.edu> <257@octopus.UUCP>
Reply-To: madd@bu-it.bu.edu (Jim Frost)
Followup-To: comp.sys.ibm.pc
Organization: Boston University Distributed Systems Group
Lines: 36

In article <257@octopus.UUCP> pete@octopus.UUCP (Pete Holzmann) writes:
|3) MCA handles multiple CPU's.
|	So does the AT bus. A kludge, true, but you can get a 68020, 32032,
|	etc coprocessor board for your AT. Even for a PC for that matter,
|	but then the 8 bit bus really starts to cause trouble! This brings
|	up the real limitation of the 'AT' bus: we need a standardized 32
|	bit extension! The MCA can handle >2 processors, so someday and
|	in some high end applications, the AT bus may be a liability for
|	this.

I have used an AT with 8 coprocessors.  And I know for a fact it would
work with 31, since the software which coordinates all of this is
designed for it.

So:  you don't need the MCA bus for even that many processors.  You
just have to break your back to make it work.  (Please don't scream at
me about how difficult some bus functions are with this setup.  It's
not the point.)

There is a point that you didn't bring up regarding the bus.  The MCA
bus handles stacked interrupts, which is something that that AT bus
sorely needs if you are trying to handle lots of I/O at high rates of
speed.  You might note that this is exactly the kind of application
IBM designed the bus for, since the PS/2 series is supposed to help
with mainframe connectivity.  And mainframes are IBM's business, not
PC's.

BTW, the MCA architecture isn't exactly state of the art.  IBM has
been doing much the same (and even more advanced architectures) for a
long time.  It's novel for a PC to have that architecture, but it's
not state of the art.

Cheers,

jim frost
madd@bu-it.bu.edu