Some friend's and I have discussed putting together a board with a MIPS
R3000 family CPU to plug into an ISA bus machine.  The board would be
designed with low cost in mind, however, should still manage good
performance.  These are the proposed specifications:

        *  MIPS R3081 CPU (20, 25, 33 or 40 MHz)
        *  16, 20 or 24 SIMM sockets for 1MB or 4MB SIMMS allowing
           64MB, 80MB, 96MB of memory using 4MB SIMMS
           (actual number of sockets depends on space left during design)
        *  ISA bus peripheral card  (the interface should be able to
           sustain 2MB/sec throughput including software overhead).

The R3081 is a full R3000 CPU and MMU and a R3010 FPU on one chip.  When
clocked at 40Mhz, it delivers 35 MIPS.  I have measured a 33MHz R3000 at
about 10 times the speed of a 486-33.

The pro's and con's of an ISA bus card rather than a mother-board for
this project:

Pro's:
        *  Cheaper (no kbd cntlr, bus interface, DMA, sockets, etc)
        *  R&D time dramatically reduced (no need to build a ISA bus
           interface that every x86 mother-board must have)
        *  The mother-board becomes an I/O co-processor, leaving more
           time for the R3000 to do real work
        *  Cheaper to buy a x86 mother-board (less than us$100) and
           this ISA card, rather than to buy this card as a mother-board

Con's:
        *  I/O would have a higher latency, because the data must also
           be copied to/from the R3000 board
        *  Must also have a x86 mother-board.

I've made some initial guesses at price:

        *  20MHz CPU without FPU, 0 RAM installed:      us$450
        *  40MHZ CPU with FPU, 0 RAM installed:         us$650

These prices are not final; the final price will depend on exactly
what parts are used and in what quantities they can be purchased in.
The above guesses are based on producing 50 boards in each production
run (most companies produce hundreds, if not thousands).  The prices
are still way lower than other companies would sell this for since
I am not charging anything for my time and there are no middle men.

Why am I doing this?  Given the above cost, one of these is much
better bang-for-the-buck than the higher priced mother-boards available
today.  I don't mind doing the R&D to make this, in much the same
way that people don't mind writing free software.

Why I'm telling you this?  The above prices are to produce at least
50 boards.  To make just one would cost about us$3000.  Also, since
I intend this to be a "free software machine", there has to be others
involved; free software is successful because there are so many
people working on it.

What I need from you?  Enough interest to justify spending the kind of
money to do a large enough production run to be economical, and
people interested in porting Linux to it when its done.

As implied above, these boards are not for sale, yet.  If you are interested
in getting involved, please send me a note; address below.

PS.  Pyramid has nothing to do with this project.  This project would not
compete in any way with Pyramid Business.  Pyramid is only my day job.
-- 
Neil Russell            (The wizard from OZ)
Pyramid Technology                      Email:  caret@pyramid.com
3860 N. First Street                    Voice:  (408) 428-7302
San Jose, CA 95134-1702                   FAX:  (408) 428-8845