[Top] [All Lists]

Response to lots of mail

To: caret@pyramid.com
Subject: Response to lots of mail
From: Jerry Callen <jcallen@Think.COM>
Date: Tue, 29 Jun 93 10:46:25 EDT
Cc: riscy@pyramid.com
In-reply-to: Neil Russell's message of Mon, 28 Jun 93 18:56:36 PDT <9306290156.AA26387@sword.eng.pyramid.com>
   From: caret@pyramid.com (Neil Russell)
   Date: Mon, 28 Jun 93 18:56:36 PDT

   Pat says:
   > Neil, are you still planning on using 32 SIMM sockets?

   I'd very much like this many SIMM sockets there.  Code for the R3000 is
   fairly large; the same program on the R3000 is about 40% (a guess)
   larger than on a x86.  Given that this system could handle more users
   than a x86 (mostly because of the SCSI thing I beleive) you want more
   memory.  Imagine 20 odd users all with X-terminals hanging off this
   thing.           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

OK, I'll stop lurking.

Frankly, I can't POSSIBLY imagine 20 odd users all with X-terminals hanging
off this thing. I don't have 20 life forms, including plants, in my entire

What ARE the goals for this project? I think there are people with vastly
varying agendas out there.

Personally, I'd like a cheap, compact system that runs some Unix variant
and has a simple expansion bus that I can build toys for. I'm mostly a
programmer, but I like to build silly little toys in my attic.  For MY
goals, Steve's proposal makes sense:

   > cpu
   > ram
   > rom
   > video
   > scsi
   > leds
   > i/o bus
   > serial i/o
   > nvram & clock

The FPU is a non-issue, since the 3051 and 3081 are pin compatible.

A 40MHz clock makes sense; that way the same board can take a cheap, 20MHz
3051, a less cheap 20MHz 3081, or a gold-plated 40MHz 3081. Builder
decides. No changes.

More SIMMs means more money for sockets, more real estate, and (if you want
to play it safe) more nasty, power-hungry driver chips. It also complicates
life if you want to implement a memory system with better page mode logic
than you get with the default 30x1 scheme using WRNEAR. For MY goals, one
bank of SIMMs that accepts 1/4/16Mb parts gives me enough flexibility.

As Steve's pointed out, he and I've been kicking this around for a while.
We've looked at an AMAZING spectrum of Rube Goldberg configurations, and we
keep coming back to: make it simple; make it cheap; don't use hard to get
or work with parts; don't get too ambitious. That way lies madness.

Steve> My I/O bus is just 8 data lines, 8 or more address lines, power, clock, 
     > 3051 family specific handshake signals.  It takes about 36 pins.  The
     > connector is like the one Apple used in the Classic for memory addon - 
     > don't need any gold fingers.  You CAN use a ribbon cable to connect to a 
     > for prototyping.  The burst rate approaches 20MB/sec.

Neil> I still maintain that having a new I/O bus that no-one has peripherals 
    > for is totally useless.

Don't forget that Steve's proposal includes a SCSI port. You can get a lot
of weird devices for SCSI these days.

It is incredibly easy to hook peripherals to Steve's bus. You can add a
floppy controller (my pet project) with essentially no glue except the
decoding. Ditto a parallel port. We're not talking rocket science to build
add-in boards. The hardware AND software are easy. This is ideal for
someone like me who wants to play with weird little toys like motor
controllers. Again, it gets back to goals.

Now you all know what _I_ am interested in. What about the rest of you?

-- Jerry Callen
   jcallen@world.std.com           (preferred)
   jcallen@think.com               (OK, too)
   {uunet,harvard}!think!jcallen   (if you must)


<Prev in Thread] Current Thread [Next in Thread>