G-15 at CMU

Athena	Bendix G-20 and CIT G-21	Univac 1108	IBM S/360-67	IBM 1401	Site Home
IBM 7040	Hybrid Lab	G-15	Bogart	DEC PDP-10	CMU Photos from Chris Hausler

G-15

This Page last updated on .

This page continues the documentation of the computers at Carnegie-Mellon University in the 1960's. You can click on the links above to learn about some of the other computers.

Hi Roy,

The first dectape I bought, in fall 1967 IIRC, was for the PDP-8 in HH53, however, when the dectape drives were installed on it that fall, us undergraduate students were told to keep our hands off of the machine. So I eventually used this tape on the Hybrid Lab PDP-9. This tape came in a cardboard box and was labeled "Micro Tape", the dectape name had yet to be assigned. I still have a printed listing of the directory of this tape :-) As I recall it cost me about $15 in the CIT bookstore. A couple years later when the PDP-10 was installed, I bought a few more dectapes for it. Now they came in round blue plastic boxes and the name "dectape" was applied to them. Looking at one of these plastic covers, there are check boxes for two possible formats, 2702(base 8) blocks of 201(base 8)12 bit words per block for PDP-5/8's and 1102(base 8) 400(base 8) 18 bit words per block for PDP-1/4/6/7/9/10/15's. (However, being cheap, I bought them untested and unformatted :-) This was before the PDP-11 came to be which had yet a third format due to it being 16 bits, but I never used a dectape on a PDP-11. But you could format a dectape on the PDP-9 and use it on the 10. These were the only two machines on which I ever used dectapes. When I went out into the "real world", for the first 10 years or so I was using Data General minis so by the time I started using PDP-11's dectapes were long replaced by floppy disks. The dectape was essentially the floppy before floppies were developed but only Digital ever used them to my knowledge. However, I believe they were developed at Lincoln Labs and originally called "Linc Tape". DEC acquired the idea when they built the "Linc 8" which was some kind of extension to the PDP-8. At least so I recall...

Regards,
Chris

Chris,

The first PDP-8 I used was located in the room adjacent to the G21 at CMU. This particular one only had a paper tape and a tty at the time, and I found it so frustrating to load everything by PT and have it crash five seconds later.

Well nightmares to me, I entered a lab in my first year at Pitt with a DEC LINC-8. Basically, this was a PDP-8 with an A-to-D convertor and some added instructions to handle the analog stuff. It was an awful computer, had lots of timing problems (remember that the PDP8 was an asynchonous logic design without a central clock), and everytime the unit heated up, it crashed. If I remember right, the DECtapes were on the PDP side of the computer (there wasn't much intelligence on the LINC side). I did see some pictures of the original LINC which was even more primitive. DEC basically viewed the LINC-8 as an opportunity to enter the real time lab business and modified the PDP8 to act like a LINC. About 3-4 years later, I worked on the PDP12, the successor to the LINC-8 which was stable and a workhorse.

Dave

Hi Dave,

Are you sure about a PDP-8 near the G-20's? What time frame was this? To my knowledge, the first PDP-8 on campus was the one in HH53. Remember HH53 was Gordon Bell's anteoffice and he was supposedly the principal developer of the 8, so this makes sense. This was also the room in which the G-15 was located. Are you sure you aren't thinking of the G-15 in this case. (I have 8x10 B&W prints of myself of photos, taken by Pat Stakem, sitting at the consoles of both the G-15 and the PDP-8 in HH53 in fall 66 :-) At least beginning in fall 1966 there was no PDP-8 in Scaife and none appeared until that aborted attempt to use one as an RJE front end to the 1108 was started and I don't think that happened until about 1969 or late 68. Maybe Bell's 8 was in Scaife before fall 66 but certainly not for long as they weren't first delivered until well into 1965. That PDP-8 which did appear in Scaife in 69 was located in the northeast corner of Scaife behind the 1108 tape drives, about as far as one could get on the fourth floor from the G-20's (but the G-20's were gone by that time and the CS PDP-10 was where they once stood). When I first arrived in fall 66 the PDP-8 in HH53 only had the ASR33 for paper tape I/O but some time that fall a high speed paper tape reader and punch unit was acquired for it. The next fall, 67, the dectapes were added and so was a single platter disk drive. That system became a graduate student project and thus us undergrads were refused access. By this time too, the G-15 had started to fail regularly and I don't believe I tried to use it more that once that fall and that would have been the last time I used it.

But of course, my brain cells have been atrophying for a number of years and all the above may be false memories, or nightmares, as the case may be...

Regards,
Chris
AN GETTO$;DUMP;RUN,ALGOL,TAPE
$$

Gordon Bell

From Chris Hausler: Jan 2009
Well, Attached are three photos, one of me in front of G-15 in HH53 and one of me in front of PDP-8 in HH53. Both of these were taken by Pat in the fall of 1966. Enjoy! More photos will be coming eventually but probably slowly. I'm going to be in Birmingham, AL a lot this spring and may not have time to put much together. I'm also going to likely be in Cumberland, MD a week in February. I've already told Pat. CSX hump yards in both places.

I've attached a third photo of me there showing me manually operating one of the G-15 tape drives.

I'm attaching a fourth photo which just shows the ASR35 which was in HH53. As there was rarely anyone waiting to use this TTY, unlike the ones in Scaife, if TTY turn time was short one could frequently get a number of runs in an evening from this TTY. I can remember early on getting quite a few chess moves done in one evening from this TTY, playing that chess playing program which ran on the G-20.

The 8 in Scaife was definitely there late probably in the time frame you mentioned - and the area was behind the 1108. It did not initially have DECtape, but did have one later and was intended for use as a front end for the 1108. The area I was referring to is approximately the same as you describe - it depended on how you travelled. I did not have much access to HH. I started school in 1966, but was in the CC starting as a high school student in 1964.

dc

Hi Chris,

I tend to agree with you. The first PDP-8 showed up in the G-15 room when Bell arrived. 1967 as I remember.

Roy

Date: Thu, 11 Mar 2010 10:16:13 -0500 (EST)
From: "J. Chris Hausler" <jchausler

earthlink.net>
Subject: Re: (Fwd) Athena - Jim Pollock

Hi Mark (and Jim!),

Here's a photo of a B250 flop (this one hung from the rear view mirror of my car for most of the 70's) and a G-15 flop. I was holding off on sending you this photo as I was going to add a PDP-8 R202 dual flop and a DDP-516 (ARPANET IMP) card and maybe others as well. But I haven't gotten around to it yet. So here's what I have. Hopefully Jim can correct or update some of our fuzzy memories on all the hardware changes he made to the Athena.

73,
Chris Hausler

Jon Orovitz sent me this in 2018

Bendix G-15 Experiences
by Jonathan Orovitz

During my first (autumn 1962) semester at Carnegie Tech I took a programming course (s205). CIT was one of very few schools offering programming courses. That summer I had learned IBM 7090 FORTRAN and assembler programming at NASA.

Although Bendix is best known for washing machines and brakes, they were also in the computer business. The G-15 was their first digital computer. Their earlier differential analyzer, an analog computer, is often seen in old sci-fi movies. When it came out in 1956 a base model G-15 cost $49.5K. That is $453K in 2018 money.

Around 1960 Carnegie Tech ordered a G-20 from Bendix for delivery in 1961. The school built a new computer center (Scaife Hall) to house it. The G-20 was a state-of-the-art second-generation, all solid-state computer in the same category as an IBM 7040 (a couple of years away).

The G-15 had been a 1961 interim “gift” from Bendix who were late delivering their new G-20. When the G-20 was installed in 1962, the obsolescent G-15 was set up a few yards away, in the user preparation area, also on the fourth floor of Scaife Hall. The user prep area had large windows, tables, chairs and keypunches. This was before Scaife
housed an IBM 1401, 7040 or RCA 301. Unlike the G-20, the old G-15 was not behind glass.

The user prep area also an IBM card sorter. Blank 80-column IBM cards were available in boxes of 2000 ($2) from the college book store. To save money some students used the card sorter to find usable discarded IBM cards. Since the G-20's GATE language ignored data after column 40, many used cards could be actively recycled by flipping them over.

Some people argue that the G-15 was the first personal computer. Its primary input device was an IBM electric typewriter. It was the size of a refrigerator and could easily be rolled around on wheels. The major argument against it was the price.

Most of the time the G-15 sat powered off, used only by a few enthusiasts. I found someone to show me how to work it and where the manuals were stashed. Soon I also got my own set of manuals from the local Bendix office. To a college student G-15's advantage was its availability while batch programs, submitted to the G-20 on punched cards, took many hours to come back. One or two runs a day were normal.

The G-15 used vacuum tubes and solid-state diodes with 2K 29-bit words of RAM on a spinning drum. When internal temperature rose above 80F the machine would fail. Some clever person drilled a hole in the front panel and inserted a meat thermometer.

Near the top center of the G-15's front was a paper tape strip and cassette reader. Just below the reader was a paper tape punch. When powered on, the computer tried to read diagnostic routines from that reader. No diagnostics - computer no go.

The G-15 was slow but useful. The G-15 had a high-level programming language called ALGO, an ancestor of ALGOL 60, CIT’s ALGOL-20, Pascal, and Delphi. Without benefit of optional magnetic tape drives, compiling a minimal (Begin-End pair or Hello World) program took 45 minutes.The ALGO compiler’s three phases were on three paper tape cassettes. Each compiler phase punched out an intermediate tape. Fortunately during the first phase the computer could detect some syntax errors as you typed. You could save the source code on paper tape and edit it later. The same could be done with the executable program.

The G-15s native machine language was complex. In retrospect I think the complexity was due to the way data was recorded and re-recorded on the drum. I never mastered the native assembly language. I only knew two people who really understood it.

Many computers of the day used six-bit character codes so the binary shorthand was octal (0-7). For some reason the G-15's literature used hexadecimal shorthand (0-9, A-F) but they called it sexadecimal. Go figure! Hexadecimal became popular when the 8-bit byte arrived a few years later with S/360.

Fortunately there were three simplified interpreted pseudo-assemblers (Intercom 500, 1000, and Algo OpCode). The Intercom pseudo assembler came in two flavors: single precision called 500 and double-precision called 1000. They differed in the way they implemented floating point arithmetic (analogous to scientific notation). Single-precision was
good for six to eight digits. With double-precision, the number of decimal digits doubled. Either version was a great improvement over a slide rule’s three-digit precision.

Lacking floating point hardware G-15's floating point software functions really crawled. Just for laughs I wrote a two-instruction program in Intercom 1000 program. It added one to a counter (a register) and branched (or jumped) back to that instruction. As soon as the program started executing I began counting to 100 out loud. When I got to 100 I stopped the program and looked at the register. It was only at 78!

Finding an error in an ALGO executable code, you could patch it with Algo-Opcode. This pseudo assembler was much like Intercom but could reference ALGO’s symbol table. After patching with Algo OpCode one could re-execute Algo programs without recompiling them.

As an E&S student I took an introductory physics course. During a physics lab we recreated a classic experiment, first performed by Galileo, timing the progress of a ball rolling down an inclined plane. While our timings were likely more accurate than Galileos, the results did not produce a smooth graph. Being inveterate wise-ass, I wrote a small ALGO
program, plotted the results and then overlaid a smoothed out graph line.

I graphed the results on the G-15's typewriter then pasted the whole thing together along with the ALGO source code and submitted the assignment. I received a severe reprimand for using a computer! We were still in the K&E Log Log Decitrig Duplex era.

In 1963 Bendix sold its computer business to Control Data Corporation. CDC continued to support the G-15 and G-20 for some time.

When I returned to Carnegie Tech in autumn 1964, for a delayed sophomore year, the G-15 had moved to the bowels of Machinery Hall where is sat largely idle. By then someone added a pair of primitive magnetic tape drives that were used as work file during compilations. This reduced Algo’s compile overhead from 45-minutes to 10 or 15. Unlike modern tape drives, these lacked built-in, real-time erase heads. Without erase heads you put up a new tape, did your business with it and, when it was eventually full, take it over to a manual degausser. When degaussing tapes it was always good practice to remove your wrist watch (and pacemaker). Mechanical watches of the 1960s did not respond well to intense magnetic fields.

Machinery Hall was an old building lacking central air conditioning. Th room in which the G-15 sat was often warm and stuffy despite a window AC unit. When the meat thermometer approached 80 deg F I opened the hinged side panels. If that didn’t help I rolled the damned thing over to the AC unit so it go the cold air blowing directly through it.

Eventually the G-15's owners restricted access to their own students and that is the end of my story.

Jon,

Some questions about your G-15 story and related. First, on the G-20, during a visit to CMU about 15 years ago I saw some photos of a G-20 console being unloaded in front of GSIA, there is evidence that it was first installed there. You say you arrived in fall 62. Was the G-20 already in Scaife at that time and was there just one of them or had both processors been delivered by then? There is also evidence that the school had an IBM 650 which I think I saw in HH53 in fall 65 but it was gone, replaced by a PDP-8 when I arrived as a freshman in fall 66. This was the same room in which the G-15 was at that time installed. Was there any evidence of that IBM 650 when you arrived?

I did not realize that the G-15 was an appeasement from Bendix for late delivery of the G-20, interesting! Your price listed for the G-15 is what I understood it to be but recently came across a document advertising new ones for a lot less, see: http://s3data.computerhistory.org/brochures/bendix.g-15.1955.102646277.pdf . I don't know the date of the document. You mention it being considered the first personal computer. Sometime in the mid 70's some trade magazine I was getting free at work celebrated the 20th anniversary of the G-15 and called it the first personal computer :-)

All I ever did with the G-15 is write a couple of small Algo programs. I remember looking at the Intercom manual but didn't ever program in it. I only got the tape drives to work once and otherwise stuck with the paper tape punch output for the intermediate steps. This was fall 66. By spring the G-15 had become unreliable and I gave up using it. I did not understand that thing about the erase heads and maybe that was my problem with the tapes. Those drives were made by Potter, the same company which made the G-20's one inch drives and I never thought much of them either although I still have a couple one inch tapes I used on the 20's. I do not remember Algo Opcode but I doubt at the time that I would have understood the G-15's weird architecture enough to use it. The G-15 was supposedly based on Alan Turing's design of the Pilot Ace.

Chris Hausler

Chris,

In the fall of 1962 Scaife was new and had lots of unused space. The G-20 installation seemed fairly recent. I wasn't shown it during an April orientation visit. I'm guessing it happened over the summer. In '62 there was only one. My best guess is that second G-20 arrived during the summer of 1964.

In autumn 1962 there was no evidence of an IBM 650. One occasionally heard oblique references to it and GAT. I'm sure it was never in Scaife.

Your undated Bendix G-15 document strongly suggests it is much later than 1956. It mentions "350 installations." Other literature says 400 G-15s were built. It also references an ALGOL compiler so I feel that brochure probably dates to 1961-63. By that time IBM had a competitive baby solid-state computer: the 1620. I used one during the summer of 1962.

By 1976 the term "personal computer" was in the English language whereas it wasn't known in 1962. BTW being a systems programmer in the 1970s I often used mainframes (after hours and holidays) as my personal computers. Potter made a respectable 1/2-inch magnetic tape drive. I used several when I ran an S/360 service bureau circa 1970. Potters were cheaper, and in some ways more advanced, than their competitors.

The joy of the G-15's pseudo assemblers were that they emulated a much simpler single-address machine. You didn't have to worry about including the next instruction address or which tracks to rewrite when.

Jon

From: <cstrathmeyer

comcast.net>
Subject:    CMU and their Bendix G-15
Date sent:    Mon, 16 Aug 2021 22:28:29 -0400

Hello Mark,

Through the randomness of Google I happened across your notes about the CMU Bendix G-15 vacuum tube computer in the 1960's.

In the summer of 1964, I was a rising high school junior and was awarded an NSF grant to attend a computer science summer school program at CMU. My project was character recognition, and I was assigned to use the G-15 (then in Machinery Hall). I remember working my way down to the room where the G-15 lived, passing gigantic dusty power transformers on casters that had
evidently been used in prior decades for power engineering courses.

By 1964, the G-15 had been equipped with the magnetic tape drives mentioned in Jonathan Orovitz's notes and I could write and compile programs in the G-15 dialect of ALGOL. I also tried writing a short routine in G-15 machine language and found that keeping all the instruction execution times and drum rotation timings in my head was not worth the effort!

I sketched out handwritten characters and recorded coordinates using a data reduction easel like a draftsman's table with big crosshairs (I recall its name was something like a Fisher Data Dibbler). Those coordinates went onto punch cards, and when eventually we found that the G-15 wasn't up to the data crunching task I was allowed to submit jobs to the 'big' G-20 in Scaife Hall. (I recall that encoding the characters required by ALGOL, like square brackets, required some gymnastics on the IBM card punch machines and if you tried to duplicate cards with that many multi-punched columns the card punch machine made some horrible unhappy noises.)

I eventually went on to a math/computer science degree at Dartmouth College and pursued a career in software, data networks, and telephone networks. I finished my career with 16 US patents in computer-telephony, so I guess the CMU/NSF program and the venerable G-15 launched me in a good direction.

--Carl Strathmeyer
   Lunenburg MA

G-15 Links

http://www.computermuseum.li/Testpage/Bendix-G15-1950s.htm
http://www.piercefuller.com/collect/bendix/index.html

The Way It Was: Tales from a life in computing - My second computer: The Bendix G-15

Here are some old G-15 computer ads.

Although I had come across that old ads site before, I hadn't spent a lot of time with it. I see the tape drives on the G-20 were made by Potter. There's a nice brochure about the Univac 1004 III showing the Uniservo VI tape drive. I remember seeing one of those Minivac 601's for sale in a local electronics store.

Regards,
Chris Hausler

Manuals at bitsavers.org: http://www.bitsavers.org/pdf/bendix/g-15/

Here are some G-15 manuals sent to me by Chris Hausler. Click on the cover page for the pdf file.

Chris wrote:

Hi Mark,

With all this talk about the G-15 I wondered whether you wanted to add any G-15 manuals to your page on the machine. Over time I have found several of them on-line but I just checked several of my links and they've all gone dead. However I had downloaded several items back when the links were still working. The only one I remember using with the machine was the Algo language manual. There is also an Algo operation manual (how to run Algo), an Intercom manual and a basic machine coding manual. I remember seeing the Intercom manual but never used it. I can send any or all of these to you, a little over 20 MB total. One link that did work was to a number of sales brochures for the G-15 at:

http://www.computerhistory.org/brochures/a-c/bendix-corporation/

but although despite coming from the Computer History Museum which one would have thought would have known about what they were talking mentions that the G-20 was a transistorized version of the G-15 which we all know is wrong.

73, Chris Hausler

ALGO - Algebraic Compiler for the BENDIX G-15 general purpose digital computer