Talk:IBM 7030 Stretch

ZZZZSTUV

In the section "Registers", address 10 $SB, there appears: "ZZZZSTUV". Can someone explain what those characters mean? John Vandenberg 05:39, 14 October 2006 (UTC)

This is how the bits of the register are identified in the IBM documentation.

• S - Sign bit
• T - Flag bit
• U - Flag bit
• V - Flag bit
• Z - Zone bit

See: Reference Manual page 36. -- RTC (talk) 06:59, 30 January 2008 (UTC)

The fastest computer in the world from 1961 until 1964.

This claim is challenged by the three instances of the Ferranti Atlas in the UK. However, such claims need precise justification, so I've not altered the text, see - . Michaelwilson (talk) —Preceding undated comment added 14:30, 6 June 2009 (UTC).

Reference columns

two columns

one column

I don't think having two columns of references helps. It does save two lines, but it breaks #1, 7, & 8. Bubba73 ^{You talkin' to me?} 01:35, 17 October 2014 (UTC)

Software list

I've added a tentative link to a COLASL article, but I won't be trying to write it immediately since to do so implies transcribing stuff from the LASL book with mathematical symbols and coloured text. If anybody feels up to it be my guest :-) MarkMLl (talk) 11:10, 30 October 2014 (UTC)

Some details

How many were built/used? Were there different configurations? Which devices were used for I/O and programming (card readers/punchers etc.) --Mopskatze (talk) 04:02, 9 May 2017 (UTC)

I recall the number being 13; various sources claim 9, and don't mention C-E-I-R, Inc, which I recall as having been one of the customers.

Like most computers, the configuration was whatever you ordered. Neither the memory size nor the peripheral devices were one size fits all. IBM offered card readers, card puinches, printers, magnetic tape drives, disk drives, etc., and you ordered as many of each model as you needed, and could always add more if necessary. Shmuel (Seymour J.) Metz Username:Chatul (talk) 01:11, 28 May 2020 (UTC)

I've found a reference to the order from CEIR; does anybody know whether it was cancelled or delivered? -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 16:46, 20 November 2023 (UTC)

References

External links modified

Hello fellow Wikipedians,

I have just modified one external link on IBM 7030 Stretch. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:

• Added archive https://web.archive.org/web/20080705164759/http://researchweb.watson.ibm.com/journal/rd/255/ibmrd2505W.pdf to http://researchweb.watson.ibm.com/journal/rd/255/ibmrd2505W.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Y An editor has reviewed this edit and fixed any errors that were found.

• If you have discovered URLs which were erroneously considered dead by the bot, you can report them with this tool.
• If you found an error with any archives or the URLs themselves, you can fix them with this tool.

Cheers.—InternetArchiveBot (Report bug) 00:39, 10 November 2017 (UTC)

It sends you to a PDF, but neither Preview on Sierra nor the "2017 release" of Acrobat Reader DC on Sierra can read it - they both say it's damaged. I went to the IEEE site instead, even though it's a paywall. Guy Harris (talk) 06:33, 10 November 2017 (UTC)

7030 vs. Atlas

There is a question posted above about the 7030 vs. Atlas. Looking into this, the 7030 slightly, but did, best the Atlas. Atlas did an add in about 1.59 micros while Atlas was 1.38 to 1.5. So I think it is safe to say that the 7030 was the world's fastest through this period. That said, it seems Ferranti's main sales point for the Atlas was that it had much higher overall throughput than the 7030 due to its fantastic channel support (512 devices!) and multiprogramming OS that was able to keep the system busy (and ran 1000 programs a day on average). In overall workload terms, Atlas did outperform the 7030. So which was the most powerful, depending on your definition, is still open to argument. Ultimately while IBM considered the 7030 to be a failure (for a time anyway), Atlas was considered a huge success and sold several additional machines until CDC came along and rained on everyone's parade. Maury Markowitz (talk) 20:06, 16 December 2019 (UTC)

Atlas did an add in about 1.59 micros while Atlas was 1.38 to 1.5. So I think it is safe to say that the 7030 was the world's fastest through this period.

Presumably the second "Atlas" in that quote should be "the 7030". Guy Harris (talk) 20:19, 16 December 2019 (UTC)

Indeed Maury Markowitz (talk) 21:34, 16 December 2019 (UTC)

Most likely the comparison should be other than just add time. As well as I know, the 7030 was used at LASL for bomb design calculations. For CPU bound problems, I/O speed is less important, and maybe also multitasking is less important. Gah4 (talk) 02:07, 17 December 2019 (UTC)

Feel free to peruse the references added to the Atlas article, which has all the metrics. Maury Markowitz (talk) 12:07, 17 December 2019 (UTC)

Any realistic comparison of performance must take into account the effects of memory size and overlap. As I understand it, the raw instruction time on Atlas was slightly faster but Stretch had better pipelining. Also, while the Atlas design allowed or a much larger memory size than stretch, the units actually shipped were smaller. Even the drums were tiny. Shmuel (Seymour J.) Metz Username:Chatul (talk) 01:06, 28 May 2020 (UTC)

As well as I remember without looking it up, the 7030 was considered a failure given its cost/performance. They had to sell them for about half the expected price. Gah4 (talk) 20:17, 12 September 2023 (UTC)

Timeline: architecture or marketing niche

Should the terms predecessor and successor refer to architecture or to market niche? I would have expected them to refer to architecture, in which case Stretch had neither a predecessor nor a successor, but the article seems to be referring to market niche. Where is the best place to ask this question in a more general context? --Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:09, 22 December 2021 (UTC)

I would suspect market niche, but maybe you can ask in WP:MOS. Gah4 (talk) 00:54, 14 September 2023 (UTC)

Designer

Why is Amdahl listed as the sole designer of the Stretch? He quit IBM because his original design was not going to be followed by Dunwell. Shouldn't we be attributing the design, if not partly, to Werner Buccholz who actually saw the project to the end? Benforeva (talk) 03:45, 1 May 2023 (UTC)

Prefetch?

The article claims that Stretch pioneered prefetch, but links a redirect to Cache prefetching; Stretch did not have a cache. Stretch did, however, pipeline instruction fetches.

The article mentions System/360 Models 91, 95 and 195, and the IBM 3090 series, as having instruction pipelining, prefetch and decoding, and memory interleaving; seeming to imply a gap during which they were not used. However, the 360/85, 370/165, 370/168, 3032 and 3033 also had asynchronous I-units.

The article probably should mention that other vendors had memory interleaving in the 1960s, e.g., CDC. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 08:59, 12 September 2023 (UTC)

The 360/91 is a favorite, or was for many years, in books on pipelined processors. The 7030 is rarely discussed. But okay, as well as I remember, the 360/91 can prefetch on three paths. That is, following two possible conditional branch paths. I suspect prefetch on a single path was not so rare by then. I don't know specifically for the 7030, though. I usually look at Blaauw and Brooks book for these questions. Gah4 (talk) 20:20, 12 September 2023 (UTC)

Prefetch input queue appears to be an article about instructio prefetching, but it may suffer a bit from x86ism; for example, is "prefetch input queue" used as a name for instruction prefetch queues in general, or is it what Intel called it on some x86 processors? Guy Harris (talk) 01:07, 13 September 2023 (UTC)

There is no right answer when trying to match up features from different processors. For x86, it is complicated by the need to put words or bytes together to make instructions. If there is a common CS name, independent of the proprietary name, we could use that. When I find my Blaauw and Brooks, I will see what they say about it. Gah4 (talk) 00:52, 14 September 2023 (UTC)

From IBM's Early Computers, page 452: The design of the CPU's instruction and lookahead units provided(sic) for a high degree of concurrency. Up to eleven successive instructions could(sic) be in the CPU registers at various stages of execution: undergoing address modification, awaiting operands from memory, waiting for or being processed by arithmetic units, or waiting for a result to be stored in memory. Since instructions could be waiting for or being processed while others are being fetched, it makes it complicated to define prefetch. In any case, I suspect much of that is new in the 7030. The 360/91 has the ability to fetch from store buffers, which is often important, and maybe the 7030 doesn't have. Gah4 (talk) 02:03, 14 September 2023 (UTC)

IMHO, the 7030 was the most successful failed product in History, and it left a deep impact on future IBM products. I certainly know of no previous computer with pipelining. In comparing the 7030 to newer machines, expect differences in nomenclature, e.g., buffer instead of look-ahead level. -- Shmuel (Seymour J.) Metz Username:Chatul (talk).

Yes, much of it went into the 360/91. "... multiprogramming, memory protection, generalized interrupt, memory interleaving, lookahead, eight-bit character, and a standard interface for I/O devices" are S/360 features that originated on Stretch.

Instruction format?

I added a link to the description of instruction formats in the manual. Should the article also show the actual formats, not just the sizes? -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 18:35, 17 May 2024 (UTC)

CDC vs. IBM

The claim that "Within IBM, being eclipsed by the smaller Control Data Corporation seemed hard to accept" is misleading. The reference for this is about the CDC 6600. The claim comes in the initial section, after an explanation that Stretch lost out to Univac's LARC that Livermore chose over Stretch (not based on actual performance, but project performance as judged by Livermore). The first CDC 6600s were not delivered until 1965 (the first customers being Livermore and Los Alamos), four years after Stretch. By then, IBM had moved on to its 360 line, although none of the new machines at that time were "supercomputers" (IBM's battle with the Justice Department was in part based on claims of unfair marketing by IBM promising that it's 360/91 and 360/95 would run circles around the CDC 6600 when in fact they did not -- or at least not to the extent IBM marketing said they would).

I think the claim needs to either moved to a later section about the impact of Stretch and what followed, or it needs to be contextualized. Specifically, it needs to be made clear that Stretch had already lost out to Univac's LARC for the initial competition but that it then held the championship for five years until CDC began the next generation of supercomputers with its 6600. To be fair, I do not think that the 360 development was primarily intended to produce a supercomputer. The importance of 360 was the concept of a machine architecture (rather than one-off designs) for a family of computer, not producing a supercomputer. Ksbooth (talk) 21:51, 21 January 2026 (UTC)

The reference was Watson's Jr.'s 1963 letter. At that point, the 7030 had been withdrawn, the 360 line was under development but hadn't been announced yet, and the 6600 had just been announced.

Presumably the letter was complaining that CDC had claimed to surpass the no-longer-being-sold 7030. Mark Smotherman's ACS timeline says that a "Project X" supercomputer started in 1961, that in June 1961 "work starts on New Product Line (NPL), which later becomes the System/360", and that on September 5, 1963 "Project X becomes NPL 604 family member", i.e. it turned into S/360 model 9x (for various values of x over time). The Model 92 was announced in August 1964, and, in November, "Model 91 announced (having less expensive core memory than Model 92 so as to compete more directly with the CDC 6600)".

So perhaps CDC's 6600 announcement, presumably claiming performance surpassing the 7030, provoked Watson Jr.'s letter. I'm not sure from the timeline at what point Project X turned into an S/360; IBM Advanced Computer Systems project § IBM and CDC says:

IBM introduced its first supercomputer, the IBM 7030 Stretch, in May 1961. They had to withdraw it from the market when tests at the launch customer, Los Alamos Scientific Laboratory, demonstrated it had very poor real-world performance. Almost immediately, IBM organized two development projects, Project X at the IBM Poughkeepsie Laboratory and Project Y at the IBM Thomas J. Watson Research Center. Project X was tasked with designing a machine that would run 10 to 20 times as fast as Stretch, while Y was to be 100 times faster.

In the spring of 1962, Control Data Corporation (CDC) announced that they had installed two computers at Lawrence Radiation Laboratory and had received a contract for a third, a much more powerful design. That new machine was officially announced in August 1963 as the CDC 6600, causing IBM CEO Thomas J. Watson Jr. to write a now-famous memo asking how it was that this small company could produce machines that outperformed those from IBM.

At a meeting in September 1963, IBM decided to shore up the high-end of what was then known as the New Product Line, or NPL. Project X was directed to implement the NPL instruction set, becoming a high-end machine in that lineup. When NPL was launched in 1964 as the System/360, Project X became the Model 92, later renamed Model 91. Eventually, about a dozen machines in the Model 90 series would be sold.

Project Y was never directed to use NPL, as it was a longer-term project aimed purely at the scientific market. Development was assigned to Jack Bertram and his Experimental Computers and Programming Group and started in earnest in late 1963. Bertram brought in John Cocke, Frances Allen, Brian Randell, Herb Schorr, and Edward H. Sussenguth, among others. Schorr developed the initial instruction set and recruited his former student, Lynn Conway, to work on a system simulator.

so, while the letter didn't provoke the creation of Project X, it may have provoked a "make it an S/360" move, perhaps to get it out the door faster. Guy Harris (talk) 23:52, 21 January 2026 (UTC)

Bibliography item from IBM Advanced Computer Systems project:

• Smotherman, Mark K.; Sussenguth, Edward H.; Robelen, Russell J. (2016). "The IBM ACS Project". IEEE Annals of the History of Computing. 38 (1): 60–74. doi:10.1109/MAHC.2015.50.

References