On 28 Jun, 2007, at 04:59, cctalk-request at
classiccmp.org wrote:
Message: 3
Date: Wed, 27 Jun 2007 15:23:32 -0700
From: "Billy Pettit" <Billy.Pettit at wdc.com>
Subject: What are the really unusual or weird computers?
To: <cctalk at classiccmp.org>
Message-ID:
<5BC121186B788C48A0EE35A16FD0D34D29C6E2 at wdscexbe01.sc.wdc.com>
Content-Type: text/plain; charset="us-ascii"
This thread has been a real disappointment. Almost all of the
responses
have been about computers using standard microprocessors - off the
shelf
components. Yes a few had non-vanilla flavored OS's, or idiotic I/O
schemes. A few were even painted different colors from PC Beige.
Then you are not reading all the replies.
But nobody got into the really weird internals that
have made the
industry
so fascinating. Go back to the real early days, like the Atlas,
that let
you build your instruction set from scratch using micro-code.
Nobody seemed
to remember that most of the late 50's and early 60's used 40 bits
as a
standard.
MOST????? STANDARD????? Rubbish! IBM 7094 - 36 bits. ICT
1301 - 48 bits. CDC 6600/7600 60 (or was it 64?) bits. CDC SC17 (not
sure exact era) - 16 bits. Elliott 903/920B/905/920C/920ATC - 18
bits. Many of the BCD machines used 4 bit words I believe. Mid 60s
ICL 1900 - 24 bits. What used 40 bits?
What about the MicroData machines with a build your
own
instructions on the fly?
Tell us more please. Microcoded or 'Extra code' ?
And then there were the ultra-strange like the G-15 -
29 bit word
size, all
instructions were modified moves through arithmetic logic or I/O
devices.
The I/O devices were actually part of the internal logic - no
channels.
Actual physical memory, access my DMA from the device or just memory
mapped I/O ?
Burroughs had some fascinating ideas on virtual memory
in the 5500
series.
Seymour Cray lived weird and unusual in most of his designs.
Several people
have developed machines to run high level languages in native mode:
ADA at
Rational, APL on the Star 100, LISP, COBOL, etc.
There's not much unusual about putting some glue logic around a $3
micro
chip. We've all done it. How about the truly weird machines?
Doesn't
anyone remember when logic didn't come in million transistor packages?
Remember? I am restoring/maintaining an ICT 1301 which has individual
Germanium transistors, wire-OR, four and gates to a PCB, one flip-
flop one a PCB, a clock derived from the timing track of the last
addressed drum store, a core store unit weighing half a ton an stores
just 2000 x 48 bit words (plus 2000 x 2 parity bits). Its got Ampex
TM4 mag tape drives (not industry standard 7 or 9 track, these are
ten track units with hubs the same design as professional audio tapes
and the 2 and 3 inch wide video tapes once used by TV broadcasters).
Its got readers and punches for 80 column cards and 5, (6 or 7?) and
8 track paper tape, its got a 600 line per minute barrel printer with
120 print positions. The mag tape transfers are done by DMA but the
rest and unbuffered peripherals, if you want to print a line on the
printer, you check to see what character is in line with the hammers,
see if you have any of them in your line, and fire those print
hammers which match, then wait until the print barrel moves on then
repeat until you have printed the whole line. Then you look up where
the paper advance mechanism currently is and select the next sprag
and release current one. If you want to page feed, release all the
sprags and count the 60 lines as they go past.
To punch a card, in software, turn the 3 phase contactor on to start
the motor, wait until its at the correct speed, activate the picker
knives to start the card in motion (sideways), look to see which if
any of your 80 columns need a hole in the first (10) row, wait until
that row is under the punch bail, fire those interposers, repeat for
the 11 row, then the 0,1,2,3,4,5,6,7,8 and 9 rows. Then feed the next
card and at the same time get ready to verify the card you just
punched. Wait until the 10 row is under the check reading wire
brushes, read them in and see if they match, repeat for all the other
eleven rows, if anything fails to verify, send the card (and the
following one) to the reject hopper, stop the punch and inform the
operator.
Come on people: there were computers long before there
were
microcomputers.
And some of them still work and have 'design council' award winning
control consoles 4 feet wide and 2 and a bit feet tall, consume 13KVA
three phase and weigh about 5 tons.
And many of them were wonderfully different and
creative.
Indeed. And some of them almost make you cry because so much more
could have been done with the same amount of electonics. My machine
has been modified to implement an index instruction. Previously all
indexing and indirection had to be done by program modification, and
even now subroutine return is done that way (see my previous e-mail).
I have one machine in 'conserved' state, unmolested, unrepaired non-
runner, and one with extra tweeks and darn right mass rewiring which
runs and I can't stop thinking about how it could be improved, yet
somehow manage to stop myself doing so. There are so many gaps in the
instruction code and spare bits in the instructions etc. The only
modification I am working on plugs into an extension port lashed up
by a previous owner. This is to capture the data from the machine
onto modern media. May replace with an RS232 interface later to drive
a teletype and/or pen plotter, and/or a parallel inteface for a
Friden Flexowriter.
Anyone got a spare plug for the i/o port of the Flexowriter, or for
an IBM keypunch (model 836) or spare patch leads for the patch panel
(4mm with a ball in them).
Anyone got a spare Ellliott paper tape reader, preferably 1000
characters per second but a 300cps unit would do. My machine
originally had two readers but somehow I only ended up with one of them.
Anyone got any spare 'ICL Standard Interface' peripherals I could
plug into the vacant port on the machine?
Anyone in the UK want an Elliott surface grinder which leaks
hydraulic fluid or want an Alba shaper (a planing machine for steel,
a brutal thing, takes out about 1/16 - 1/8 at every stroke, comes off
bright red). Or a really old pillar drill, large capacity, looks
prehistoric. All three phase of course.
Again in the UK, anyone got any punched card trays, the steel type,
preferably with the racks too, but the trays only would be a help.
Willing to pay (reasonable) price for any of the above.
Roger Holmes.
Classic computer collector, classic car collector, machine tool
collector/user (for the prior mentioned hobbies), and for a job,
programmer of CAD and graphic software and printer/plotter drivers
for Apple computers.
Billy