7090 vs. 7094 etc.

35 views
Skip to first unread message

Larry

unread,
Feb 17, 2001, 11:32:15 PM2/17/01
to
IBM did an upgrade of the 7000 series; e.g. 7070 went to 7074, 7090 went to
7094, and I think 7040 went to 7044. I don't remember if there was a 7084 or
not. Does anyone know what was involved in these upgrades? For example, what
advance in technology was involved, did the cycle times improve, etc.? Also,
has anyone come upon good sources for on line manuals of the old machines
(probably pdf format)? I'm thinking of 1st and 2nd generation stuff here.


ric...@panix.com

unread,
Feb 18, 2001, 9:44:18 AM2/18/01
to

Having worked on both 7090s and 7094s back in the dark ages, I will give
it a shot. As far as manuals go, pdf hadn't been invented yet.
Both were second generation. I have no idea of cycle times except that
they were glacially slow by today's standards. The 7094 was faster than
the 7090.

The 7090 had 3 index registers controlled by a 3 bit field in the
instruction. (general purpose registers hadn't been invented yet
either.) IX1, IX2 & IX4. If more than 1 bit was on, the 2 or 3 registers
were or'd together. The 7094 had 7 index registers, and one only was
selected by the same 3 bits as a number 0-7. 0 = no index. There was a
switch on the front panel that made the 7094 index registers work as the
7090 had for backwards compatability. That was the biggest difference
from the programmer's viewpoint. I believe that the 7094 introduced
external channels while the 7090 did PC style i/o.

Most 709x's had no disks. Main i/o was tapes (round) with 800 bpi as
the "high" density. The programmer could select high or low density,
a switch on the channel defined hi/lo as one of: 800/556, 556/200, and
(I think) 800/200. They also had a modified 407 accounting machine as a
card reader (row binary) and "console" printer. Normal jobs were tape
in, tape out with card reading and printing/punching done on a seperate
machine (usually a small 1401).
--
Rich Greenberg Work: Rich.Greenberg atsign worldspan.com +1 770-563-6656
N6LRT Marietta, GA, USA Play: richgr atsign panix.com +1 770-321-6507
Eastern time zone. I speak for myself & my dogs only. VM'er since CP-67
Canines:Val(Chinook,CGC,TT), Red & Shasta(Husky,(RIP)) Owner:Chinook-L
Atlanta Siberian Husky Rescue. Adopt a homeless Husky. Asst Owner:Sibernet-L

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to list...@bama.ua.edu with the message: GET IBM-MAIN INFO

Bob Halpern

unread,
Feb 18, 2001, 10:33:50 AM2/18/01
to
I'll add a little bit to this. Most of the in/out for the 7094 was using
offline unit record equipment, including separate units, or a 1401, etc.
Simpson & Crabtree of IBM at NASA, in Houston, implemented a direct
connect from a 7040 to do the I/O. That was the 1st Direct Couple. The
idea was good, and IBM had planned a 1401/1410 and 1305 disk between for
spooling but the mathematical models showed it would be very slow. So
IBM
took the direct couple concept to the Western Data Processing Center at
UCLA, and with UCLA and IBM Los Angeles Scientific Center staff made the
Direct Couple that was to be shipped. An IBM CE named Dave Bottles came
up with the direct connection/interrupt between the 7044/7094 and the
IBSYS operating system was modified to use it. Direct Couple was born
and worked.

The team that did Direct Couple evolved into the ASP system for the 360
where a 40 would spool for a 65 (or whatever). This evolved into JES3.
JES3 went to the Thousand Oaks center for some time, and then moved back
to be with the JES2 folks under Crabtree. Simpson went to Amdahl.

Art Walters of IBM at WDPC stayed with JES3 until it moved back east.
Jobi Citron did the mathematical models that showed the 1305 disk was
too slow (500 lb head assembly). I worked on the 7740 front end that
fed communications to the 7044 from remote STR and 1050 devices. Art was
on the 7740, then RAX, then ASP, then JES3.

It was the remote STR services that brought computing to universities
all over the western United States. That is what led to the first
ARPAnet
being done at UCLA (using the 3 7094 nodes on campus), then eventually
to the Internet. WDPC was associated with the Graduate School of
Business
at UCLA, and Dr. George Brown, who headed WDPC for a while, sponsored
the
Educom conference to the University of Colorado, in Boulder, in 1966
that set off the funding for everything.

And the history went on. Ivan Southerland was the ARPA head for the
first ARPAnet, and he eventually teamed up with Dave Evans at the
University of Utah to make Evans/Southerland Graphics, the folks with
the
first virtual reality system. (I told Dave I thought he was nuts,
because
E/S graphics was located at the bottom of U of Utah, and he had to move
from his hilltop office with a 20+ foot glass view of the entire Salt
Lake Valley. He had been the head of the computing department).

Meanwhile people like Leonard Kleinerock got involved and the internet
was
born. The first IMPs (node computers) were the Honeywell 516 military
version because of ARPA (The military version cost twice as much as
the commercial version and was half as reliable). Honeywell got it
because when DEC started out, they could not keep the interface machines
going. I ended up calling on DEC in Maynard over that, but it was too
late).

It has been fun, in retrospect, to realize what I was a part of (not
a major role by any means, but more than Al Gore). I was also the
technical consultant to the Educom conference and got to meet a bunch
of really smart folks, some of whom envisioned where we are today with
the internet. One of the names that I remember most was Sam Alexander,
who received a medal for his part in computing. He was raring to go
with it, and I had to explain that some of what he wanted could not be
done (yet) as the technology had to evolve further. This was the days
when 2000 bps modems on a dial network were a "really big deal".

Edward J. Finnell, III , Ed

unread,
Feb 18, 2001, 6:13:48 PM2/18/01
to
I only did Fortran on the 7094(36bit words). Helped some guy with valance bonding
of ammonia. Think his whole dissertation was 6 900 hour runs on a spare up
at George C. Marshall. Kids running around today can flick their wrists on
an SGI and do the same thing in 5 seconds(or less).
Edward(Ed) J. Finnell, III
Enterprise Systems/Proj. Mgr.
url:www.ua.edu

George Fogg

unread,
Feb 18, 2001, 7:16:00 PM2/18/01
to
>Having worked on both 7090s and 7094s back in the dark ages, I will give
>it a shot. As far as manuals go, pdf hadn't been invented yet.
>Both were second generation. I have no idea of cycle times except that
>they were glacially slow by today's standards. The 7094 was faster than
>the 7090.

Sure bring back memories Rich. I remember the drum for storage, the fortran
library was on tape and forever being hammered on, a card reader that read
(it seemed) 1 card per 5 seconds, a punch that jammed 4 times a shift, and a
console panel that look like it came straight out of a cartoon with it's
switches, lights, dials, and rocker switches. And to boot, a 36 bit, base 8,
octal machine.

George Fogg

Neil Stutz

unread,
Feb 18, 2001, 9:25:23 PM2/18/01
to
I worked on the 704, 7090, and 7094. Here's what I recall.

The 704, which was not solid state, had a 12 microsec cycle time. The
cycle time for the 7090 was 2.18 microsecs; the 7094 was 2 microsecs.
From a programming perspective, I recall the 7090 and 7094 as being
identical. Most ops took 2 cycles, some took 1 cycle, and some took
more than 2. The online printer for all 3 machines was indeed a
modified 40? accounting machine. If I recall correctly, it printed
132-character lines at 75 lpm. The online card reader was not a
modified 40?, and it read cards, all 12X80 punches per card, at 75 cpm.
All 3 machines had a 32K memory of 36-bit words. The machines were only
"octal" in the sense that users used octal notation just as 390 users
use hexadecimal notation.

I don't recall there being more than 3 index registers on the 7094;
OR'ing of the registers would occur if an index address other than
01,2,4 was specified. Was there a 7094-II which had 7 unique index
registers? Drum or disk was optional.

Larry

unread,
Feb 18, 2001, 11:24:54 PM2/18/01
to
I realize pdf wasn't around at the time of the 7090, but I have found a 7090
manual in pdf form on the web, as well as 7040, 650, 305 and 7030 (stretch).
I'd like to get more if possible.

One nice thing I noticed in the 7090 manual is that they tell you exactly
how some instructions work. For example the multiply operation is explained
in detail as a bunch of shift/add cycles with the adds skipped if a low
order bit is zero.

The fact that IBM upgraded several of the 7000 series makes me think they
might have applied some new technology to the upgraded manuals; e.g. faster
circuitry in some critical places.


Larry

unread,
Feb 18, 2001, 11:26:39 PM2/18/01
to
Wasn't it HASP that Simpson and Crabtree, and a couple of others, did
instead of ASP?


Anne & Lynn Wheeler

unread,
Feb 19, 2001, 11:57:53 AM2/19/01
to
"Larry" <mrb...@cedar.alberni.net> writes:

> Wasn't it HASP that Simpson and Crabtree, and a couple of others, did
> instead of ASP?

yep, random HASP ref.

http://www.garlic.com/~lynn/94.html#18

Simpson went to white plains (made ibm fellow about 76) and had a
project called RASP (note there were a couple different RASP's in that
time frame) before going to dallas w/Amdahl (made amdahl fellow in
fall of '79). His RASP had somewhat the flavor of the current
incarnation of gnosis (keykos).

My wife was the "catcher" in g'burg for JES3 ... before going to POK
to be in charge of loosely-coupled architecture. originated
"peer-coupled shared data" in pok ... original basis for IMS
hot-standby and then parallel sysplex. also had some affect on
trotter/3088.

last time we ran into crabtree, he was in atlanta in charge of
mid-range system application software.

random ref:
http://www.garlic.com/~lynn/2000f.html#68
http://www.garlic.com/~lynn/2000f.html#69
http://www.garlic.com/~lynn/2000f.html#70
http://www.garlic.com/~lynn/2000f.html#71

my first share was spring '68 in houston ... & of course HASP people
were there.

hasp started a tradition of sign-a-long nights at share. I
have an old Share Songbook from the mid-70s.

--
Anne & Lynn Wheeler | ly...@garlic.com - http://www.garlic.com/~lynn/

Metz, Seymour

unread,
Feb 20, 2001, 12:22:59 PM2/20/01
to
The 7094 had 7 index registers from day one, but you had to use the leave
multiple tag mode (LMTM) instruction before using the extra 4. AFAIK there
were no programming differences between the 7094 and the 7094-II. The drum
was a leftover from the 709; I never saw a shop that had one.

I never saw a shop use the printer as anything but a clock or a console
printer; it was just too slow.


Shmuel (Seymour J.) Metz

Metz, Seymour

unread,
Feb 20, 2001, 12:22:57 PM2/20/01
to
Ah, yes, the good old RAMAC disk drive. I had hoped to forget it, in all of
it's incarnations. Suffice it to say that once IBM announced the 1301 the
RAMAC-based drives lost their luster.


Shmuel (Seymour J.) Metz

> -----Original Message-----
> From: Bob Halpern [SMTP:B...@CPUPERFORM.COM]
> Sent: Sunday, February 18, 2001 10:33 AM
>
> I'll add a little bit to this. Most of the in/out for the 7094 was using
> offline unit record equipment, including separate units, or a 1401, etc.
> Simpson & Crabtree of IBM at NASA, in Houston, implemented a direct
> connect from a 7040 to do the I/O. That was the 1st Direct Couple. The
> idea was good, and IBM had planned a 1401/1410 and 1305 disk

----------------------------------------------------------------------

Metz, Seymour

unread,
Feb 20, 2001, 12:22:55 PM2/20/01
to
They say that the memory is the first thing to go. I can't remember the
second thing.

There were actually 3 machines in the series: 7090, 7094 and 7094 II. The
7094 introduced a lot of new instructions, and the code that I've seen used
those a lot more heavily than the extra index registers. The transition
between 3 and 7 index registered was not controlled from the console, but
via the instructions EMTM and LMTM. The 7094 used the same I/O channels as
the 7090. Most shops I was aware of had a 1301 or 1302 disk drive.

The 7090 I/O didn't look remotely like the PC; perhaps you're thinking of
the 7040 and 7044, which had 704 style I/O, but even that didn't really look
like the PC.

The card reader and printer were separate boxes; only the printer used 40x
technology. The tape density wasn't controlled by switches; there were
separate models (II, IV and VI) of the 729.


Shmuel (Seymour J.) Metz

> -----Original Message-----
> From: ric...@PANIX.COM [SMTP:ric...@PANIX.COM]
> Sent: Sunday, February 18, 2001 9:44 AM
>
> The 7090 had 3 index registers controlled by a 3 bit field in the
> instruction. (general purpose registers hadn't been invented yet
> either.) IX1, IX2 & IX4. If more than 1 bit was on, the 2 or 3 registers
> were or'd together. The 7094 had 7 index registers, and one only was
> selected by the same 3 bits as a number 0-7. 0 = no index. There was a
> switch on the front panel that made the 7094 index registers work as the
> 7090 had for backwards compatability. That was the biggest difference
> from the programmer's viewpoint. I believe that the 7094 introduced
> external channels while the 7090 did PC style i/o.
>
> Most 709x's had no disks. Main i/o was tapes (round) with 800 bpi as
> the "high" density. The programmer could select high or low density,
> a switch on the channel defined hi/lo as one of: 800/556, 556/200, and
> (I think) 800/200. They also had a modified 407 accounting machine as a
> card reader (row binary) and "console" printer. Normal jobs were tape
> in, tape out with card reading and printing/punching done on a seperate
> machine (usually a small 1401).

----------------------------------------------------------------------

Bruce Black

unread,
Feb 20, 2001, 12:22:58 PM2/20/01
to
"Metz, Seymour" wrote:
>
> They say that the memory is the first thing to go. I can't remember the
> second thing.

the second thing is the sex drive, but luckily you can't remember what you
wanted it for.
--
Bruce A. Black
Senior Software Developer for
FDR, CPK, ABR, SOS, UPSTREAM, FATS/FATAR
Innovation Data Processing
Little Falls, NJ 07424
973-890-7300
personal: bbl...@fdrinnovation.com
sales info: sa...@fdrinnovation.com
tech support: sup...@fdrinnovation.com

Metz, Seymour

unread,
Feb 20, 2001, 12:50:29 PM2/20/01
to
I can remember what I wanted it for, but not its capacity or seek time.


Shmuel (Seymour J.) Metz

> -----Original Message-----
> From: Bruce Black [SMTP:bbl...@FDRINNOVATION.COM]
> Sent: Tuesday, February 20, 2001 11:27 AM
>
> the second thing is the sex drive, but luckily you can't remember what you
> wanted it for.

----------------------------------------------------------------------

Reply all
Reply to author
Forward
0 new messages