riscy
[Top] [All Lists]

Design Criteria for motherboard

To: riscy@pyramid.com (Mips 3000)
Subject: Design Criteria for motherboard
From: Bill Broadley <broadley@neurocog.lrdc.pitt.edu>
Date: Tue, 29 Jun 1993 01:52:38 -0500 (EDT)
In the design of the motherboards we have to consider that we have a finite
amount of engineering talent, money, board space, and time.  Complexity
had a direct effect on all of the above.

To maximize (performance+reliability)/price for the motherboard we should 
minimize chip counts, board surface area, and things that would be cheaper 
in volumes we can't produce.  

I think the inadequacies of the average 486 motherboard are driving this
project.   The most important of these being the cpu (we are replacing),
and the ISA bus (which we are augmenting).

I believe that way to much is making it onto the board.  The only things
that should make it should satisfy one of three requirements:  
        High enough bandwidth to me limited by ISA (appoximate 5 Mb/sec ?). 
        Need to address memory over 16 MB (isa can't)
        Much cheaper to put on the motherboard then on a ISA card.
        Necessary to debug/boot the card.

The reason that NOTHING else should be one the is because we have limited
resources:
        Board space is expensive and finite.  
        For each thing we add to the motherboard adds to: cost, design time,
assembly time, debugging time, probability of errors.   
        We aren't paying for an engineer's time (thanks!!!), but they have a 
finite amount of time.  I'd rather see the big things get all the attention 
rather then serial ports, sounds, parallel, Dsp's, etc.  The integrations
of all the possible features sounds non-trivial to me.

        I consider designing a integrated cpu+memory system+isa+scsi+
video quite an achievement under our budget/manpower.  Entire corperations
make there living on just video or scsi.  Maybe I have a misconception on
how hard it it to make an accelerated video card.

        If we put every thing that doesn't meet the above on isa cards
we protect ourselves against bugs, make the design simpler and get the
boards quicker, cheaper, and maybe use isa cards we already have (GASP).

        I want the best designed, bug free, reliable motherboard as possible.
        
        On a micromint 8088 (BEFORE the ibm pc was a standard) we had to
buy an ISA card to support a IBM pc compatible keyboard. (It's wasn't a pc
compatible when I built it.)

        There are tons of IDE+floppy+2 serial+parallel isa cards out there
probably going for around $25.00 a piece (you can turn off ide if you wish)

        Wouldn't the above add more cost if integrated into motherboard? At 
what advantage? Take a non-trival amount of engineering time?  

        Maybe a secondard project to put a keyboard controller, 16552,
parallel, floppy on a ISA card would be useful (so we can get all we want
on a single isa card.)  

My interpretation of the above. 
        Video on board, video bandwidth could easily swamp ISA bus, not to
mention I wouldn't want it eating up 2 MB or so of the addressable 16 MB
of ISA address space.
        SCSI on board, disk bandwidth can easily be a large fraction of 
ISA bandwidth (if not 100%), it's cheap to add relative to ISA card.

The above 2 leave the ISA very free, so it can do the slower things well
like it was designed to.

        Ethernet on ISA, not trivial to design (there's alot of bad ones out 
there right?), we have C source to support commercial ISA ethernet cards, max
bandwidth is around 25 % of the almost empty ISA bus.  Not everyone needs
one (ok I'm biased I don't need one).  In a year or so ISDN or FDDI may
become more popular.  

        Parallel, serial, keyboard, dsp, sounds: Easily done on isa, fairly low
bandwidth.  Exceptions it maybe impractical to debug/boot a board without
a serial port/keyboard.

        Especially with the increasing completity of the video subsystem
think we should minimize the motherboard to cpu+video+scsi.  Maybe we can find 
a cheap vesa driver and just buy $200 S3 805 cards, and then put more into
the motherboard.

        I don't think we should maximize features for $700, I think
we should minimize the performance of a mips 3000 for the minimum price.  
Even $50 cheaper (reuse your ide supercard) and 1 month sooner could
make a big difference to prospective customers.

-- 
Bill                                    1st>    Broadley@neurocog.lrdc.pitt.edu
Broadley@schneider3.lrdc.pitt.edu <2nd  3rd>                 Broadley+@pitt.edu
Linux is great.         Bike to live, live to bike.                      PGP-ok


 

<Prev in Thread] Current Thread [Next in Thread>
  • Design Criteria for motherboard, Bill Broadley <=