On 2007-11-25, Jon Elson wrote:
DoN. Nichols wrote:
On 2007-11-24, Jon Elson wrote:
Another insane project was building a 32-bit bit-slice computer.

[ ... ]

Interesting. I remember considering making a Motorola 6800 CPU
from 2-bit bit-slice modules (for a hoped increase in speed), but never
got around to starting that project.

Lots of wire-wrapped things with 6800, 6802 and 6809 CPUs.

Yeesh, that was the Signetics/Intel 3002 series, with their
insane local branching scheme, where your program listing had to
be a 2D grid, with routines writhing all over the map like snakes!

Perhaps it is just as well that I did not start collecting
hardware to try that then. :-) And at that time, the 6800 was my only
view of how a CPU should be configured.

I actually got it running at the
blinding rate of 8 MHz for 2-register operations and 6 MHz for
3-register.


Pretty good for wire-wrapped. I wonder how much it could have
been boosted by more careful design of the routing of the signals?

Not much. This was really old-school Schottky TTL, there was a
limit on how much they could put in one chip without thermal
problems.

Before the LS series, then.

Of course, I still have some plain pre-Schottky TTL just plain
"SN7400"s and similar. (Then again -- I also still have the Altair
680b. :-)

The registers were external to the ALU chips, and you
had to decode the register addresses. I could have made it un
faster with faster register chips, but they would run even
hotter, and be a lot less capable. The AMD register chips had 3
ports, 2 for reading one for writing, and there wasn't much
alternative. 55 ns access time, I seem to recall, sounded
blindingly fast at the time!

Yep -- considering what I was usually working with which was
doing amazing things if it reached a 150 nS access time. :-)

[ ... ]

Close -- CP/M.
Yes, of course, how could I have forgotten that. I spent a LOT
of time working with it, even rigged up a VERY early Memorex 10
MB Winchester drive to it, also has a 12" vector-writing CRT
with a light pen, and a Honeywell 600 LPM drum printer, also 800
BPI 9-track mag tape for backup.

At first, I did not have *any* OS for the Altair 680b -- until I
wire-wrapped an interface for a digital cassette drive and burned
routines to read and write Motorola Hex format data to/from it and added
commands to the monitor ROM to use those. Everything had to live in
1720 EPROMs (256x8 IIRC, and slow enough so the 680b was clocked down to
500 KHz instead of the native 1 MHz for the CPU. And the 6800 seemed to
do a little more at 1 MHz than the 8080 at 2MHz -- did things on each
edge of the clock pulse, instead of just one edge.

The SWTP 6800 I ran on floppys (both 8" and 5.25") for quite a
while, and was finally replaced with the SWTP 6809, using the same disks
and controller cards.
then I picked up an IMI 5MB hard disk with a controller whose
external interface looked a bit like SASI, but wasn't quite it. I first
wrote drivers for DOS-69 (rather CP/M like, except that it didn't have
PIP as a command and was 6.3 filename format -- I think that CP/M was
8.3 like early MS-DOS. However, since DOS-69 (and its predecessor
DOS-68) did not have a subdirectory structure, a 5MB drive got awfully
cluttered, so I took the time to write drivers for OS-9 (unix-like OS
for the 6809), and OS-9 was quite happy with the disks. I added another
of those to max out that controller, then added two 27 MB MFM drives
with a SASI (pre-SCSI) controller, and wire-wrapped an interface for
that. That worked quite well until it got replaced by my first unix
system, the Cosmos CMS-16/UNX which was based on an 8MHz Motorola 68000
CPU.

[ ... ]

An impressive project, anyway.

it would have been impressive if I'd ever gotten it running.
I had 2 MB of Memorex 3rd party static RAM memory out of IBM
370's at work that were scrapped. I was going to build an
interface between that memory and the 32-bit CPU, which would
have allowed me to actually run a program on the thing. But,
the microcoding was a total nightmare, partly because the tools
I had were a bit primitive. It took me a couple days to write
the code to do the simplest operations, like add 2 numbers
together and write the result in another register. The last
thing I did was get a simple 32x32 - 64 bit multiply working.
Next would have been divide, and I think that's about where the
project stopped.

O.K. About where it started to get really complex -- especially
with primitive development tools.

Then, it became possible to buy a DEC MicroVAX-II CPU piece by
piece from brokers, and I never looked back! I still pull out
the huge wire-wrapped boards for visitors to marvel at.


I moved from the Altair 680b to the SWTP 6800, and then the SWTP
6809, (the last finally running with DOS-69 and OS-9 at the flick of a
switch) before I started picking up Sun workstations and servers,
starting with a Sun 2/120, and up through a current pair of Sun Blade
1000s and a Sun Fire 280R.

I got a SGI Iris 2020 off the loading dock at work. It almost
booted up, you could look around in the file system, etc. I got
some help on the net and determined the graphics engine was bad,
and bought all the boards out of a German guy's system for $100.
His graphics engine worked, and the thing came up and ran their
OS and nifty demos like the "flight simulator". It ran for
about 2 years and then the graphics engine blew again. I sold
all the guts for $200 to a broker. This machine was so far
beyond obsolete it wasn't funny. 68020, I think.

Hmm ... Well, I went through the 68000 (on the Cosmos
CMS-16/UNX), then the 68010 (AT&T Unix-PC followed by Sun 2/120) before
I finally got to the 68020 with the Sun-3 family.

My VAX just died this year, after 20 years of operation, and a
number of upgrades. It is also mighty far out of date, 0.9
MIPS, 4 MB of memory.

I have a friend who might have been able to sell you the parts
to get it running again. But he strips machines and sells parts to
dealers, so the prices would probably not be too friendly. :-)

I had one last application running on it
for the last 7 years, an energy/environment monitoring system
that has a couple LCD displays around the house that show time,
inside and outside temp, humidity, etc. and also logs a whole
bunch of info on furnace and air cond operation to a file every
15 seconds. I finally migrated the code and interface over to
my server PC. Every night it summarizes the day's data. Used
to take 3 minutes on the VAX while the displays froze. On the
PC it takes 0.7 seconds, so the clock display never misses a beat.

Isn't it amazing how much faster today's systems are. I tend to
forget -- until I have to dig up the older ones to extract something for
someone -- in particular the AT&T Unix-PC/7300/3B1, which was 10 MHz
68010. It seemed pretty fast compared to the v7 unix on the 8MHz 68000,
but now that I am used to Ultra-SPARCs that is a different matter.

Out of curiosity -- were you running VMS or a unix on the VAX?

Enjoy,
DoN.

--
Email: | Voice (all times): (703) 938-4564
(too) near Washington D.C. | http://www.d-and-d.com/dnichols/DoN.html
--- Black Holes are where God is dividing by zero ---