Talk:IBM 700/7000 series

7070 as 650 replacement?

I'd be interested in seeing some backup for the claim "The IBM 7070, IBM 7072, and IBM 7074 were designed to provide a "transistorized IBM 650" upgrade path. They replaced the drum memory with core memory, but were not instruction set compatible with the 650 (so a simulator was needed to run old programs)." The 7070 did have a 10-digit decimal word length, same as the 650, but that is the end of the similarity. I attended one talk in the 1959-1960 time frame where someone from IBM presented the 1620 as the replacement for the 650. I have no doubt that IBM sold some 7070's to large 650 shops that wanted more capacity, but the 7070 was a much more expensive machine. And I never heard of anyone simulating the 650 for production purposes on any machine. Computer time was considered more valuable than programmer time back then, and my recollection is that 650 programs that were still needed were rewritten for whatever machine was leased as a replacement. Anyone have a different recollection or documents?--agr 22:14, 28 May 2006 (UTC)

I have an article that was written by Bob Bemer, and was posted on his website until he died and some casino took over the URL, that explains how the 7070 was originally planned as a unified upgrade path to replace both the 650 and 705. Because it was incompatible with both it was not that well liked (several incompatibilities resulted from engineering to management communication problems with managment making the decision to go with a more 650 like machine when the programmers were recommending a more 705 like machine). Ultimately IBM resolved the issue by also building the 7080, which was compatible with the 705.

As Bob Bemer's website is no longer available and the casino site that took the URL has blocked archiving, it is hard to get copies of this article (titled BIRTH OF AN UNWANTED IBM COMPUTER). Let me know if you want a copy of the HTML file. -- RTC 18:15, 7 June 2006 (UTC)

Yes, I'd like to see it. You can follow the link from my user page for my e-mail address. This sort of thing needs to be preserved. I wonder if the computer history museum would host it?--agr 19:09, 7 June 2006 (UTC)

Bemer's site is back, including the article; I've linked to it in the "External links" section. Guy Harris (talk) 23:30, 25 December 2013 (UTC)

The 700/7000 series is not a series

Is the definition of a series all the machines built in some time period using the technology of that time period and the naming conventions of that time period? No, series, or families, share architectural features and you can say useful things that apply to all of the machines in a series. Like the IBM 1400 series.

Saying that all these machines were built during the same time period and that they were obsoleted by the machines that followed them is a tautology. Other than that, the article is divided by architectures and it is those architectures that define families, that define series.

This article should be deleted, replaced by new articles, one for each architecture listed in this article. New articles like IBM 702/705/7080 series. 69.106.254.246 02:13, 27 September 2006 (UTC)

"Series" is the term generally used for these machines, see for example, http://www.piercefuller.com/library/ibm7000.html. Your definition is a more modern one and results in no small measure from IBM's problems developing software for so many different architectures. One reason to keep the article intact is to make clear to the reader how different these machines were from a programming point of view and how they each influenced the next-generation System/360 architecture.--agr 17:50, 28 September 2006 (UTC)

I find it slightly strange, but not enough to say that it should be deleted, until other articles are ready. I suspect that I believe there should be separate articles for the scientific and commercial lines. Again, I am not ready to write them. Gah4 (talk) 08:04, 11 November 2021 (UTC)

You are answering a question from 15 years ago.  Stepho  talk  09:08, 11 November 2021 (UTC)

I suppose so, but it seems to still apply. While I was reading the article yesterday, I was surprised to see tube and transistor machines together. I suspect that there could still be reason to make changes, related to the wide range of machines discussed. Gah4 (talk) 08:42, 12 November 2021 (UTC)

Tubes and transistors isn't a problem. Think of the original Intel 8086 from 1978 vs a modern i7. Both are clearly from the same series but the i7 is vastly faster, addresses more memory, has more bits per word, etc, etc. Many computer architectures got reimplemented in a later technology.  Stepho  talk  09:12, 12 November 2021 (UTC)

You can run 704 software on a 7094 II and 705 software on a 7080. I would support splitting the article into 36-bit machines, 10-digit machines, 7010 and 702/705/7080, but would oppose finer divisions.

Would you remove the 360/85 and 360/195 from IBM System/360 because the used integrated circuits? --Shmuel (Seymour J.) Metz Username:Chatul (talk) 15:12, 12 November 2021 (UTC)

"Scientific" should include 701

While the 704, 709, 7040, 7044, 7090, and 7094 have much in common (e.g., floating point, instruction format) that the 701 did not share, it's misleading to call their architecture "scientific," since the 701 was also a scientific computer. IBM's own website (https://www.ibm.com/ibm/history/exhibits/701/701_intro.html) calls it "IBM's first commercially available scientific computer." Isolating it in its own architectural category is OK; the problem can be fixed by coming up with another name for the scientific category. — Preceding unsigned comment added by Rochkind (talk • contribs) 23:45, 3 July 2019 (UTC)

I don't think that it is misleading to say that the 704, 709, 704x, and 709x had scientific architecture. The issue is whether or not the 701 belongs with them. In my view, the 701 was the first in the scientific line and the 702 was the first in the commercial/besiness line. Bubba73 ^{You talkin' to me?} 01:33, 4 July 2019 (UTC)

In the 1950s and 1960s, scientific computers were synonymous with binary calculations and storage and commercial computers were synonymous with base 10 calculations and storage. Which leaves the 701 as a scientific computer. The 360 broke new ground by having both.  Stepho  talk  10:20, 4 July 2019 (UTC)

Also characters for the business/commercial computers. (Asside: UNIVAC LARC was decimal but a scientific computer.) Bubba73 ^{You talkin' to me?} 17:50, 4 July 2019 (UTC)

I'm going to change the article to call the 701 the "first scientific" and the others the "later scientific." Thoughts? Rochkind (talk) 14:41, 4 July 2019 (UTC)

I agree with that. Bubba73 ^{You talkin' to me?} 16:33, 4 July 2019 (UTC)

Or, perhaps "scientific" and "advanced scientific"? Rochkind (talk) 16:48, 4 July 2019 (UTC)

I'd stick with "later"; that leaves just "scientific" and "commercial" categories, with all the binary machines in the former category. The 701 -> 704 instruction set transition was significant, with the addition of floating point and index registers; that should be mentioned. It looks as if the rest of the scientific line, from the 704 to the 7094, had at least some level of compatibility.

See the UNIVAC 1100/2200 series for a page that starts out discussing a bunch of incompatible machines (UNIVAC 1100/2200 series#Vacuum tube machines not mutually compatible) and then goes into the subsequent line of compatible machines (UNIVAC 1100/2200 series#UNIVAC 1100 compatible series); we might want to do something similar with the scientific series, although the "second series" starting with the 704 didn't turn into a long-lived compatible series, unlike the series starting with the UNIVAC 1107. Guy Harris (talk) 20:08, 4 July 2019 (UTC)

I looked into it, and the 701 was intended for scientific calculations. However, it had a different architecture from the other scientific computers, whose architecture was common to the group. So I agree with the current six-way breakdown of architectures. However, rather than calling them "scientific architecture", maybe reword it something like "architecture of the scientific group". Bubba73 ^{You talkin' to me?} 14:38, 5 July 2019 (UTC)

Possibly include this content

FORTRAN Assembly Program (FAP) was a macro assembler for the IBM 709, 7090, and 7094 computers of the 1950s and 60s. Its pseudo-operation BSS, used to reserve memory, is the origin of the common name of the "BSS section", still used in many assembly languages today for designating reserved memory address ranges of the type not having to be saved in the executable image.

I am going to turn that article into a redirect since it doesn't seem likely it can ever carry a proper article. Up to you guys whether this is worth adding to this article. (note that the claim of origin is contradicted by .bss) Ham Pastrami (talk) 09:03, 17 February 2008 (UTC)

Actually I did a partial merge. Sorry I couldn't clean up the presentation, but I think it'd be better to leave that to the experts. Ham Pastrami (talk) 09:10, 17 February 2008 (UTC)

Image copyright problem with Image:IBM7040.jpg

The image Image:IBM7040.jpg is used in this article under a claim of fair use, but it does not have an adequate explanation for why it meets the requirements for such images when used here. In particular, for each page the image is used on, it must have an explanation linking to that page which explains why it needs to be used on that page. Please check

• That there is a non-free use rationale on the image's description page for the use in this article.
• That this article is linked to from the image description page.

This is an automated notice by FairuseBot. For assistance on the image use policy, see Wikipedia:Media copyright questions. --18:08, 14 September 2008 (UTC)

"Price" of machines that were rented or leased but not sold?

There is talk of the price or cost of these machines, but they were never sold, only rented or leased. —Preceding unsigned comment added by 200.77.49.166 (talk) 08:12, 18 November 2010 (UTC)

"Price" no longer appears in the article; "cost" does, but machines that are rented or leased have a "cost" just as machines that are bought do. Guy Harris (talk) 23:33, 25 December 2013 (UTC)

Timeline

Any objection to my sorting the lines of the "Timeline" section to make it an actual timeline, that is, in chronological order? At the moment it's a list of model numbers which happens to mention their introduction dates. 206.205.52.162 (talk) 21:11, 4 September 2012 (UTC)

Seeing no objection, this is done. 206.205.52.162 (talk) 16:03, 12 September 2012 (UTC)

Are dates when announced or when first shipped

Are the dates intended to be when they were announced (put on sale) or when they were first shipped. 'Announced' I think would be better (if IBM had them working before announcing them?) but the 7040 is noted in 1963 (when it shipped) rather than 1961 when it was announced. We should add "announced" or "shipped" to the date column or qualify each date ? With the 7030 it was shipped in 1961 but it's not clear when 'it' was announced - so perhaps first shipment would be better to use. - Rod57 (talk) 03:02, 11 January 2016 (UTC)

Didn't they pre-announce vapor-ware by ridiculous intervals before the consent decree that made them stop that? Dicklyon (talk) 04:38, 11 January 2016 (UTC)

I would prefer that both dates be given if known. If only one date is given, it should be clearly labeled as announced or First Cusomer Shipment (FCS). Shmuel (Seymour J.) Metz Username:Chatul (talk) 18:20, 11 January 2016 (UTC)

NACA vs. NASA

About the caption to the photo, NASA didn't exist until 1958, so in 1957 it was still National Advisory Committee for Aeronautics (NACA). Bubba73 ^{You talkin' to me?} 00:56, 21 April 2014 (UTC)

7095?

I recently found manuals at http://bitsavers.org/pdf/ibm/7095/ about an unannounced successor to the 7094. Should the article mention it? Shmuel (Seymour J.) Metz Username:Chatul (talk) 16:36, 8 March 2019 (UTC)

If we can figure out from the documents on bitsavers enough to say about it, yes. It *looks* as if it may have been a "last gasp" add-on to the 709x series, with more main memory (18 bit physical addresses), double indexing, supervisor/user mode and memory protection (perhaps inspired by the CTSS add-ons), and support for "NPL", by which I think they mean "System/360", peripherals. Perhaps the intent was to provide something for customers who needed more than what a 7094 could provide but who weren't yet ready to switch to S/360 (even to an M65 with the compatibility feature). Guy Harris (talk) 20:30, 12 November 2021 (UTC)

What software to mention?

IBM 700/7000 series#Scientific architecture (704/709/7090/7094) has a section IBM 700/7000 series#Scientific architecture (704/709/7090/7094)#FORTRAN assembly program foor FAP, but not for any other 709x software. Shouldn't there be sections on at least

• COMTRAN
• FAP under IBSYS
• FORTRAN II
• IBCBC
• IBFTC
• IBJOB
• IBLDR
• IBMAP
• IOEX
• IOCS Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:55, 8 March 2019 (UTC)

References

BSS

When I linked the first reference to BSS to its Wikipedia article I did not notice that the second reference was already linked. The pseudo-op BSS originally just allocated storage in-line; it had nothing to do with storage sections in the object file, since they didn't exist when BSS and BES were first defined. The Wikipedia article describes both the BSS section and the pseudo-op as it was originally defined.

"Commercial architecture"

What exactly is the "commercial architecture"? This term should be clarified for a reader (or perhaps a separate article should be created?) --Yuriz (talk) 22:27, 3 August 2021 (UTC)

Commercial architecture means it was optimised for decimal arithmetic - primarily for the finance industry. As opposed to scientific architecture which was optimised for binary arithmetic - primarily for physics research such as calculating ballistic trajectories for missiles.  Stepho  talk  22:35, 3 August 2021 (UTC)

By that definitions, there are three commercial architectures:

1. 702, 705, 7080
2. 7070, 7072, 7074
3. 7010

The 7070/72/74 architecture is optimized for ten digit decimal arithmetic; the others are optimized for variable length decimal arithmetic. --Shmuel (Seymour J.) Metz Username:Chatul (talk) 12:20, 4 August 2021 (UTC)

Manuals?

Would it be appropriate to add reference manuals for the various 70x/70xx machines. If so, should they be, e.g., given with <ref>...</ref>, listed under external Links, listed in a separate section? Shmuel (Seymour J.) Metz Username:Chatul (talk) 10:31, 2 November 2022 (UTC)

Yes, it would be appropriate. If they're used as references for claims in the article, give them with <ref>...</ref>, otherwise list them under "External links". Guy Harris (talk) 18:12, 2 November 2022 (UTC)

Done, as a definition list, except that I couldn't find a 709 manual. I lumped 7090 and 7094 under a single label and I included |ref= for each manual to allow for later use of {{sfn}}. --Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:53, 4 November 2022 (UTC)

Video and 703

I've just added a link to a video about the 701 (the last of the external links). It mentions a 703 that was used only internally at IBM for scientific and engineering purposes. Bubba73 ^{You talkin' to me?} 03:11, 12 April 2024 (UTC)

Someone on Reddit said this about the 703: "The IBM 703 was a unique NSA model, so the module in the video is almost certainly not from that system. " I seem to remember hearing that somewhere reliable. Bubba73 ^{You talkin' to me?} 03:18, 12 April 2024 (UTC)

That would be Ken Shirrif, in this post, and I'd consider him a pretty reliable source. Guy Harris (talk) 03:27, 12 April 2024 (UTC)

I found this book (looks like a great reference), here is a PDF, (first 30 pages only) it is a legitimate book. It says that the 703 was a one-off computer for NSA and nothing else is known. Should the 703 be included in the article? Bubba73 ^{You talkin' to me?} 03:34, 12 April 2024 (UTC)

A few more pages are at Google Books: https://www.google.com/books/edition/First_Generation_Mainframes/B9l9DwAAQBAJ?hl=en&gbpv=1&printsec=frontcover Bubba73 ^{You talkin' to me?} 03:49, 12 April 2024 (UTC)

Since it is a documented, factory made system, I would include it. We don't know if it is a scientific or commercial variant, so it will have to be in it's own section with just a single paragraph saying "Delivered to NSA - nothing else known".  Stepho  talk  04:08, 12 April 2024 (UTC)

The book in question popped up in my search, too. (The categorization of 700/7000 models is similar to the one in IBM 700/7000 series § Architectures.)

On the other hand, in an issue of IEEE SSCS News, there's an article, "Computer Architecture and Amdahl’s Law", by Gene Amdahl, a well-known IBMer, in which he says:

Elaine Boehm and I determined that we had to make an outstanding demonstration to win approval. We came up with the idea of a tape sorting program. The IBM 703 was a sorter-collator, a fairly modestly priced machine sold to the US Treasury. The expected price for the 709 we estimated to be at least two or three times that of the 703. We programmed the sort and found that it performed so much faster than the 703 that the cost of sorting on the 709 was less than on the 703. This demonstration tipped the balance and the I/O Channel development was approved!

and this IBM Journal of Research and Development paper, "The Architecture of IBM’s Early Computers", one of the authors of which is Werner Buchholz, another fairly well-known old IBMer, says that

At the same time, considerable effort was spent on developing an auxiliary tape sorting and collating machine, tentatively called the 703, which was intended to relieve the main computer of much of its tape-handling burden. ... The 705 also proved to be sufficiently better than the 702 at handling tape that the plan to announce the specialized 703 tape sorter-collator was dropped.

so we have two notable IBM old-timers saying it's a tape sorting and collating machine.

Then again, as a certain young girl said in a Taco Bell commercial and subsequently in 6.02*10^23 reaction GIFs all over the Intertubes, "Porque no los dos?" Perhaps either the machine was designed for the NSA and they told the Treasury Department about it and they wanted one, too, or the machine was designed for the NSA and the cover story, used later, was that it was sold to the Treasury Department. Guy Harris (talk) 04:11, 12 April 2024 (UTC)

Or perhaps more than one project used the number 703?

The NSA do a lot of work on text ciphers, so I can see the value of a machine dedicated to maintaining/sifting/sorting those text files, quite possibly based on the commercial architecture. Pure conjecture on my part of course.  Stepho  talk  06:16, 12 April 2024 (UTC)

And, straight outta Ft. Meade, we have "Machines in the Service of Cryptanalysis", which says:

Next Spring a whole new vista in data handling will open when we receive the IBM 702 Tape Processing Machine and the 703 File Maintenance Machine and ancillary equipment. While this equipment can compute, it is different from a computer. The rest of our computers work most efficiently on problems which.allow small input, large processing, and small output of answers. The 702 complex will work well on problems with large input, large processing and large output. This should make it possible for us to give problems such as traffic analysis the high speed high volume processing which they have.never had.

So the NSA seems to indicate that there was a 703 that might attach to a 702. Perhaps, as per the Buchholz et al paper, they planned on releasing it outside the NSA, but found that the 705 had rendered it obsolete. Guy Harris (talk) 06:43, 12 April 2024 (UTC)

Seems like I've gone into search-like-a-squirrel mode:

1. https://homepage.cs.uiowa.edu/~dwjones/arch/notes/09ias.html says 703 was decimal base - making it likely to based on the 702.
2. https://imgur.com/gallery/LcvDwrY has a pic of the memory and it looks like 8-bit - excellent for 7-bit BCDIC text in 702 style but not good for 36-bit words in 701 style.
3. https://books.google.com.au/books?id=WqrJkVLxonkC&pg=PA90&lpg=PA90&dq=%22ibm+703%22&source=bl&ots=ir0Yu-J7nY&sig=ACfU3U2lPr-V0Md4mHPYbZAGUXUxprBh0g&hl=en&sa=X&ved=2ahUKEwjpq6e7i7yFAxXTxTgGHYbZBzI4HhDoAXoECAMQAw#v=onepage&q=%22ibm%20703%22&f=false says "The IBM 703, never sold commerically, was a special purpose high-speed sorting machine, using magnetic tapes. It was delivered to NSA. Later, in October 1957, the first large transistorized computing machine made by IBM was delivered to NSA." Implying that the vac-tube 703 was superseded almost immediately.
4. http://www.edwardbosworth.com/CPSC2105/Lectures/Slides_06/Chapter_02/RealNumbers.pdf says 703 used packed decimal.
5. https://ipwatchdog.com/wp-content/uploads/2019/08/Dibble.pdf says "specialized tape sorting and collating machine" (ref to be followed up: [Bashe ei al., 1981] C. J. Bashe, W. Buchholz, B. V. Hawkins, J. J. Ingram, and N. Rochester, "The Architecture of IBM's Early Computers," IBM Journal of Research and Development, 25(5):363-375, September 1981.)
6. https://www.softwarepreservation.org/projects/LISP/conference/iwoleas88/Padget-ThreeUncommonLisps.pdf says Lisp was on the 703 and 704 (page 40) as part of the lineage for Lisp/VM.
7. http://www.bitsavers.org/magazines/Computers_And_Automation/195606.pdf says "IBM 703 (Electronic Data Processing Machine for File Maintenance)".
8. https://cntaware.org/wp-content/uploads/2019/12/117stewart.pdf says "From Site startup in 1951, Savannah River scientists and engineers had only limited access to computers: a Card Punch Calculator 7 was acquired in April 1953; an IBM 650, in 1955; and an IBM 703, in January 1962." Possibly a hand-me-down?
9. https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2022CN000190 says he learnt FORTRAN on the 703 in the late 1950s. Possibly mistaken for the model 7030.
10. https://archive.computerhistory.org/resources/access/text/2019/03/102785430-05-03-acc.pdf Gordon Bell says (when asked about a special SORT instruction) "We tried just that in the IBM 703. It turned out that such a special purpose machine (it was planned as a tape sorting machine) had nearly all the necessary elements of a general purpose machine. ^ So why bother? Build the general purpose machine, and do with it what you will, like tape sorting." which matches to 709 comparison above.
11. https://www.governmentattic.org/16docs/NRO-SIGINTsatStory_1994.pdf says the "703 and UNIVAC 1103 were the first commercial digital computers to use magnetic-core storage." Ref "Encyclopedia of Computer Science, p. 1488."

Of course, take each with a grain of salt - model numbers are easy to get wrong.  Stepho  talk  08:15, 12 April 2024 (UTC)

"The Architecture of IBM's Early Computers," IBM Journal of Research and Development I mentioned that paper above, without linking it; I've fixed that. It's the one that says that an IBM 705, with its better I/O subsystem, could do the sorting/collating as well as a 703. Guy Harris (talk) 19:19, 12 April 2024 (UTC)

The best source for this is probably the book IBM's Early Computers, but my copy is boxed up right now. Bubba73 ^{You talkin' to me?} 14:34, 12 April 2024 (UTC)

but my copy is boxed up right now The Internet Archive is your friend. My copy of IBM's System/360 and Early System/370 Systems is in storage, so I've used the Internet Archive's copy. IBM's Early Computers is also available to borrow, at https://archive.org/embed/ibmsearlycompute00bash. I checked it out; the 703 is mentioned in a couple of footnotes:

M. E. Femmer, June 1968: interview by L. M. Saphire. The tape sorter-collator, for which machine type number 703 had been reserved, was a plugged-program machine using barrier-grid cathode-ray tubes for its memory.

and

P. E. Fox, October 1967: interview by L. M. Saphire. Although the barrier-grid tube was not used on IBM's commercial computers, it was used on the File Maintenance Machine, at times called the IBM 703 and later known as the 770 when contracted to the National Security Agency.

I've now returned it. The pages containing the references to the footnotes may give more detail. Guy Harris (talk) 19:32, 12 April 2024 (UTC)

Annnnd we have A History of Computing in the Twentieth Century, also available on the Internet Archive, with some mentions of the 703. Pages 410 and 411 speak of an NSA "request for the design of a machine that became the 703", speak of the 702 as having been sold to the NSA as well as several companies, and say that the 703 was built "In response to the requests of the National Security Agency" and that "It was a special-purpose machine that was intended to carry out a high-speed sort based on tape input to about 1000 words of cathode-ray-tube storage". Guy Harris (talk) 20:07, 12 April 2024 (UTC)

Then I'm not so sure that it belongs in the article with the 700 series. Bubba73 ^{You talkin' to me?} 21:51, 12 April 2024 (UTC)