Talk:List of transistorized computers

The RCA 501 computer

A close friend of mine who worked for RCA in Camden in the 1950's & 60's in the first business applacation of the 501 would like to create a wiki site on the 501 computer and it's history. Is there someone who can help him with this? Let me know. The Davison Library (talk) 23:28, 20 October 2015 (UTC)

External links modified

Hello fellow Wikipedians,

I have just modified one external link on List of transistorized computers. 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/20110930122620/http://www.dvq.com/docs/misc/an_uyk-1.pdf to http://www.dvq.com/docs/misc/an_uyk-1.pdf

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

This message was posted before February 2018. After February 2018, "External links modified" talk page sections are no longer generated or monitored by InternetArchiveBot. No special action is required regarding these talk page notices, other than regular verification using the archive tool instructions below. Editors have permission to delete these "External links modified" talk page sections if they want to de-clutter talk pages, but see the RfC before doing mass systematic removals. This message is updated dynamically through the template {{source check}} (last update: 18 January 2022).

• 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) 04:15, 3 January 2018 (UTC)

IBM 360?

Does the IBM 360, with its IBM Solid Logic Technology, count as a transistor computer? Bubba73 ^{You talkin' to me?} 23:54, 15 March 2018 (UTC)

Is a solid logic module a discrete transistor? No? --Wtshymanski (talk) 04:06, 16 March 2018 (UTC)

IBM's Solid Logic Technology (SLT) of the System/360 used discrete transistors and at least intially no integrated circuits so I think it belongs in this section. Many references refer to the third generation of computers as using ICs and it seems to me they are refering to monolithic ICs. I don't think anyone would consider SLT a hybrid IC nor the early s/360 an integrated circuit based computer. Perhaps the answer is to remove the term "hybrid" from the lede and footnote the IBM entry. Tom94022 (talk) 07:32, 16 March 2018 (UTC)

See for example, this photo Tom94022 (talk) 16:26, 16 March 2018 (UTC)

What is a hybrid IC, then? And how is it different from what IBM did with SLT? Sure, not "monolithic" but there were integrated circuits long before "monolithic" ICs came along, that's why they have their own special name. You could pull a fistful of transistors out of any of the discrete transistor computers and turn them into, oh, say, an AM radio or an audio amplifier...but the SLT modules were only usable as logic gates. --Wtshymanski (talk) 00:10, 17 March 2018 (UTC)

"What is a hybrid IC, then?" A PC board, possibly with epoxy poured on it?

What I see as significant about monolithic ICs is that the transistors are connected as part of the semiconductor fabrication process, rather than as part of the circuit-board wiring or PCB fabrication process. Putting a bunch of individually-fabbed transistors on a board and calling it a "hybrid IC" doesn't strike me as much different from putting a bunch of individually-fabricated transistors on a board and calling it a "circuit board"; fabricating a single die with multiple transistors on it does strike me as significantly different. Is there some reason to consider pre-monolithic-IC technologies such as SLT significantly different from, for example, discrete transistors in cans? Guy Harris (talk) 19:00, 17 March 2018 (UTC)

Yep. That's why IBM did it. And that's why they are different from discrete transistors. Each hybrid had a designated function - a bag of transistors could be made into anything.--Wtshymanski (talk) 19:48, 17 March 2018 (UTC)

None of the transistors in a transistorized computer are in a bag, they are on boards performing specific functions just like an SLT module and can no more be made into a radio than the discrete transistors in an SLT module. SLT is just a packaging advance and really doesn't make the early s360's into integrated circuit computers. Tom94022 (talk) 20:35, 17 March 2018 (UTC)

Yes - what makes the choice about how to use individual transistors (and diodes and resistors and capacitors and...) - put them all on the circuit board, or put small collections of them on circuit boards and then put those collections on other circuit boards - at the same level of significance as putting the components onto a single die? Guy Harris (talk) 20:52, 17 March 2018 (UTC)

IBM SLT (?) card from the 1960s

I don't know enough about the hardware to say. But when I took Intro to CS a long time ago, we covered the history of computers. IIRC, the IBM 360 was an example of a second-generation (transistor) computer and the 370 was an example of a third-generation computer. Bubba73 ^{You talkin' to me?} 23:35, 17 March 2018 (UTC)

Correction - I got my old textbook and it lists the IBM 360 as a third-generation computer. Bubba73 ^{You talkin' to me?} 02:17, 18 March 2018 (UTC)

Within the last two weeks I bought the card pictured. It was described as an IBM SLT card from the 1960s. I believe that is true since the card is so similar to the first photo in IBM Solid Logic Technology. Are these discrete transistors? (Added: that article describes them as discrete transistors.) Bubba73 ^{You talkin' to me?} 00:17, 18 March 2018 (UTC)

It looks like an SLT board, but not with SLT (square things with a metal lid) on it. Some S/360 peripherals are built with SMS, the previous generation technology. I will guess that it is needed for driving signals at higher current than SLT can drive. Though even for that, there are not enough pins. Gah4 (talk) 20:27, 16 April 2018 (UTC)

Thanks. Bubba73 ^{You talkin' to me?} 23:13, 16 April 2018 (UTC)

The transistors on an SLT module are not individually packaged and cannot be removed from the module without destroying them.They are permanently attached to a commmon substrate which is then packaged and attached to an etched circuit board. The functional unit is the module, not the individual transistor. Any electronics hobbyist of a certain age, on the other hand, will recall the joy of unsoldering discrete transistors from some computer board and making neat things out of the discrete transistors thereby removed. Theres an IBM brochure online that talks about these hybrid integrated circuits. [1]. The little round cans in the Bubba73's photo don't look like the square modules in the IBM brochure. --Wtshymanski (talk) 03:10, 18 March 2018 (UTC)

Are they transistors? The have the Texas Instruments logo, then "291" an then "6708B". Bubba73 ^{You talkin' to me?} 03:22, 18 March 2018 (UTC)

Well, they probably aren't SLT modules which are shown as square. The case does look like a discrete transistor case. Do they only have 3 leads each? That would make it a transistor...probing with a voltmeter and looking for junctions would tell the tale. "291" is probably a "house number" and not a JEDEC part number...when you buy enough transistors, you put your own part number on them. And if "6708B" is the date code, then they date back to early 1967. It's probably not a 2N291, which is in an oblong TO22 package - none of my data books are old enough to show it, but on-line it's a PNP germanium transistor. --Wtshymanski (talk) 03:35, 18 March 2018 (UTC)

Yes, they do have three leads each. The board part is very similar to the photo at IBM Solid Logic Technology and the labeling is similar. That one has "4619 A 5105YB". Mine has "04196 F 8808 YB". Bubba73 ^{You talkin' to me?} 03:47, 18 March 2018 (UTC)

Reading the aforementioned article, mine sounds more like "Solid Logic Dense". Bubba73 ^{You talkin' to me?} 03:57, 18 March 2018 (UTC)

I don't know of anybody in the industry that regarded SLT as monolithic, and IBM certainly distinguished them in, e.g., the 360/195.Shmuel (Seymour J.) Metz Username:Chatul (talk) 19:47, 18 March 2018 (UTC)

,No, a hybrid integrated circuit isn't a monolithic integrated circuit - that's why they have different names. --Wtshymanski (talk) 16:49, 19 March 2018 (UTC)

The online IBM brochure on SLT includes the following statement:

In an SLT module (below), the squares (1) are the tiny chip transistors and diodes which are attached at junctions on the circuit pattern (2). The large dark areas (3) are resistors.

IMO "tiny chip transistors" are discrete transistors making the early S/360s transitorized computers. I'd also note there is no RS for the lede's lumping of hybrid circuits and monolithic integrated circuits into one bucket separate from discrete. The lede could just as easily lump discrete and hybrid into one bucket. In the absense of an RS we could strike it. Tom94022 (talk) 21:13, 14 April 2018 (UTC)

From this ([2] p. 424, [3] Introduction/Transistor and diode fabrication) it looks like SLT was a combination of several (two at least) monolithic ICs into one module. Not a pure monolithic IC, but for me it's close enough. 89.25.210.104 (talk) 00:02, 23 April 2018 (UTC)

Presumably you meant "it looks like SLT was a combination of several individual transistors, diodes, and passive components into one module" - they weren't monolithic ICs on the substrate. I view it as significantly different from monolithic IC technology. Guy Harris (talk) 00:25, 23 April 2018 (UTC)

There is nothing in either of yr referecnces that has even one monolithic IC in an SLT device. Pugh Chapter 8 which is yr first reference makes it clear that even as of 1964 IC meant monolithic IC and IBM was critized for and concerned by the lack of monlithic ICs. The images in yr second reference show no multi-leaded chips which would be characteristic of monolithic ICs. Tom94022 (talk) 06:50, 23 April 2018 (UTC)

Criticized, yes, but in Pugh they successfully defended the criticism. Monolithic ICs weren't up to the reliability and production scale needed for S/360 when they would have been needed. On the other hand, it isn't so obvious that they couldn't have transitioned to monolithic based SLC within the lifetime of S/360 production. Though the storage keys for the 360/91 were built with 16 bit SRAM chips, that must have been packages somehow. Gah4 (talk) 07:55, 23 April 2018 (UTC)

Agree,it would have been foolish to commit early S/360 to ICs, so doesn't that make them second generation, transistor computers? 19:08, 23 April 2018 (UTC)

But if you read Pugh, they are way ahead of the previous technology. Glass passivated, aluminum metallization wires to ball bonding spots. If I had to give it a number, I would say 2.5th generation. The diodes have two on a chip, too! Gah4 (talk) 21:34, 23 April 2018 (UTC)

Perhaps 2.5 but not 3 :-). We agree that the early S/360 computers did not use monolithic integrated circuits required for a third generation computer. Certainly the early S/360's were very advanced in their electronic packaging; however the logic in SLT was mainly individual semiconductors. This article is a list of computers that mainly use discrete transistors (as opposed to transistors in monolithic ICs) for logic and that is exactly what we have here. They should be listed here with a footnote describing their technology and/or a link to the SLT article. BTW the SLT article notes MST was the variant that used monolithic ICs. Since we have an integer taxonomy where would u list them? Tom94022 (talk) 00:41, 24 April 2018 (UTC)

How about a separate section on either this page or the IC computer page, explaining the difference and listing appropriate computers? (Are there other hybrid circuit computers?) That probably makes more sense than a separate page, but that would work, too. Gah4 (talk) 01:53, 24 April 2018 (UTC)

May I suggest a separate section on electronic packaging in the main article, Transistor computer, would be appropriate? The RS's for the section would be Pugh and Gregg; I'll put something there in the next day or two. I don't know of any other computer's using hybrid circuits but IBM is so significant to the computer industry that it seems their transistor computers ought to be on this list. Tom94022 (talk) 05:55, 24 April 2018 (UTC)

I just looked at Discrete transistor and no, SLT is not discrete transistors, which implies separate packaging. Yes it is make with individual transistor chips, but they are not packaged before being put on the ceramic blocks. They are glass passivated, such that they don't need a hermetic seal, but otherwise are not packaged. Gah4 (talk) 06:01, 24 April 2018 (UTC)

• This is semantics but the SLT transistors are "individually package transistors" per Discrete transistor since the glass passification counts as package - not a standard package as listed in the Discrete transistor section but that is an ommission of that article's section and shouldn't limit this article. FWIW the term "discrete" does not appear in any of the RS's on computer generations that I've reviewed. It appears to have been added herein to distinguish between the transistors of monolithic ICs and microprocessors. We can ignore this distinction without much substance herein or I suppose I could add a sentance to the Discrete transistor section linking to SLT and see if it is accepted there. Tom94022 (talk) 16:43, 24 April 2018 (UTC)

There seems to be an actual article: List of integrated circuit packaging types. That says about packaging: to allow easy handling and assembly onto printed circuit boards and to protect the devices from damage. Glass passivation does protect against damage, but doesn't allow for easy handling. Today, one might install them directly on PC boards, but that is somewhat recent. They are enough smaller than the usual IC not to allow easy handling. The ceramic chips are meant to allow assembly onto printed circuit boards. Gah4 (talk) 17:27, 24 April 2018 (UTC)

From Flip Chip (PDP module): the early use of hybrid integrated circuits allowed DEC to market the PDP-8 as an integrated circuit computer. Gah4 (talk) 18:12, 24 April 2018 (UTC)

Thanks, your pointer to Flip Chip (PDP module) further makes the point that S/360 should be added to this list. Again we are dealing in semantics - Some people in the 60s may have described DEC and IBM systems as using "hybrid integrated circuits" but that is not the current definiton of IC. FWIW the PDP-8, 9, and 10 are all used FlipChip and are listed in this article - should they now be stricken? Tom94022 (talk) 21:04, 24 April 2018 (UTC)

It seems that DEC also used the FlipChip name for some things that weren't. Notice, for example, the picture on that page that has ordinary three wire packaged transistors on a PC board. And so now we have more machines that people might recognize (as in not military internal that were never seen outside). I think a separate section with a description of the differences would be about right. Gah4 (talk) 22:05, 24 April 2018 (UTC)

Also, the storage protection keys on the 360/91 are built from 16 bit SRAM chips, and the cache on the 360/85 uses 64 bit SRAM chips. Gah4 (talk) 17:09, 25 April 2018 (UTC)

That makes the 360/91 and 360/85 either an early 3^rd generation computer or a transistor computer with some ICs and they could be listed in either article or not at all.

We already have links to sections and articles; this article and it's main article link to Discrete transistor section which in turn links to SLT article. I suggest that only a footnote is necessary when the early S/360's are added; something along the lines of "Used discrete transistors in hybrid SLT devices."

I take it you agree that early S/360s should be added and the only remaining quesion is whether the links are sufficient or should we add a section on packaging to either this article or the main article or both? If so, why don't u try and write the section? Tom94022 (talk) 17:05, 26 April 2018 (UTC)

Beside the IBM SLT (360, 1130) (at least till 1965) there was also couple of less known (military, one-of-a-kind?) hybrid IC computers: L-2010, micropac and compac II (RCA micromodule) 89.25.210.104 (talk) 11:28, 24 April 2018 (UTC)

The PDP 8, 9 & 10 already listed in the article used a form of hybrid circuit. As apparently did the not listed PDP-7 Tom94022 (talk) 21:04, 24 April 2018 (UTC)

Actually although DEC advertised the PDP 7 and 8 as having "WITH FLIP CHlp™' INTEGRATED CIRCUITS" I can see nothing about the early Flip Chip series that is integrated or hybrid - they look like ordinary pcbs. The M Series Flip Chip apparently did use ICs

If you look at the picture in flip chip, you see plastic packaged IC looking objects. But it seems that there are also things that DEC continued to use the name for, that aren't. Gah4 (talk) 23:46, 26 April 2018 (UTC)

The black plastic lumps in the DEC R107 picture in flip chip look nothing like any IC package of the 1960s. Note it has 7 discrete transisters for seven invertors and it's schematic shows the black things typically resistors and diodes. Maybe we can call the black things hybird circuits but they sure aren't monolithic ICs. I went thru every schematic for the R Series and found a few of these "hybrid circuits" so if these are ICs then we should remove the PDP's from this list. On the other hand, if they are not ICs then neither is SLT and we should add the early S/360s to this list. Tom94022 (talk) 06:09, 27 April 2018 (UTC)

There are supposed to be FlipChip, which they might actually be, even with only diodes. By the way, SLT uses chips with two diodes. It seems that FlipChip was DECs competition for SLT, but they weren't as good at it. That is, they are supposed to be hybrid ICs. Then later they used FlipChip name for boards with ordinary three wire transistors on them. I don't know which DEC machines have FlipChip that are not hybrid ICs, and which ones are hybrid circuits. Gah4 (talk) 06:55, 27 April 2018 (UTC)

Yes, as Flip Chip (PDP module) says:

The modules were called Flip-Chips because early versions of some of these modules, for example, the R107 module shown, used hybrid integrated circuits built using flip chip mounting of individual diode chips on a ceramic substrate. Some boards containing flip chip modules were etched and drilled to allow those modules to be replaced by discrete components. At some points during production, conventional discrete components may have replaced these flip-chip devices, but the early use of hybrid integrated circuits allowed DEC to market the PDP-8 as an integrated circuit computer.

When DEC began to use monolithic integrated circuits, they continued to refer to their circuit boards as "Flip-Chip" modules, despite the fact that actual flip chip mounting was not used.

so the DEC term "Flip Chip" doesn't indicate that flip chip mounting is involved. Guy Harris (talk) 01:54, 27 April 2018 (UTC)

I don't think in current usage the early Flip Chip modules would be called anything more than pwb assemblies and until the 1968 M Series used in the 8/I I think all the Flip Chip modules used discrete transistors. The black lumps on the [R107] and S111] contain only diodes and resistors. I suppose we can call them hybrid circuits. I went thru every schematic for the R Series and found a few of these "hybrid circuits" so if these are ICs then we should remove the PDP's from this list. On the other hand, if they are not ICs then neither is SLT and we should add the early S/360s to this list. Tom94022 (talk) 06:09, 27 April 2018 (UTC)

According to soemtron pdp7flipchips, ICs were used from 1967 (pdp-8/i).

Earlier hybrid IC (diode/resistor) was used: ed-thelen/Historical Notes/Technology (PDP-8 faq).

This link may be handy to determine the type of IC: wylie The first integrated circuits (pictures are clickable, hybrid section unfortunately contains only SLT).

--89.25.210.104 (talk) 23:46, 27 April 2018 (UTC)

References

CDC Cyber 70 and Cyber 175 - should they be here?

CDC Cyber says of the Cyber 70 series that it was a minor upgrade from the 6000 series. I assume that means that they still used discrete transistors. It says that the Cyber 172, 173, and 174 used ICs whereas the 175 used high-speed transistors. So should the CDC Cyber 70 and 175 be listed in this article? Bubba73 ^{You talkin' to me?} 22:43, 18 March 2018 (UTC)

I thought the Cyber 17x were 7600 architecture, though that might have been considered a minor upgrade by some from the 6600. Gah4 (talk) 20:31, 16 April 2018 (UTC)

The CDC Cyber article says that the Cyber 70 was based on the 6600 and the 170 was based on the 7600, and I believe that is correct. (I used a 6400, a 70/74, and two in the 170 line.) Bubba73 ^{You talkin' to me?} 23:19, 16 April 2018 (UTC)

Poll on adding early S/360

Wtshymanski without further discussion reverted the addition of selected S/360 models to the list of transistorized computers. It appears from the discussion here and at the System/360 there is an overwhelming consensus to add them to this article.

Poll

Should selected early S/360 computers using SLT be added to this article?

• In favor: The discrete diode and transistor chips in the SLT module are not ICs Tom94022 (talk) 16:37, 16 April 2018 (UTC)
• In favor: Individual transistors directly attached to an SLT module, rather than put in a can with three wires coming out of it, strike me as discrete transistors; they're fabricated independently of the other transistors, rather than, as in a monolithic IC, fabricated together with the other transistors on the IC. Guy Harris (talk) 17:18, 16 April 2018 (UTC)
• In favor The transistors in a module are fabricated separately, not as part of the module. Subsequent bonding to the module does not change that. Shmuel (Seymour J.) Metz Username:Chatul (talk) 15:28, 2 May 2018 (UTC)
• Just wrong methodology We don't get to make up items by voting. What do references say? An SLT module is not a discrete transistor. An SLT is not a monolithic IC,it's a hybrid or think-film IC; the world existed before Bill Gates invented the IBM PC in 1981. This article is about computers that use discrete transistors. --Wtshymanski (talk) 21:08, 2 May 2018 (UTC)

• Polling may help us arrive at a consensus and in the absence of concensus will help an administrator determine if there is a rough consensus. Tom94022 (talk) 06:54, 4 May 2018 (UTC)

* Undecided - my old textbook lists the 360 as a third-generation computer, but IBM's statement in the book about the 360 and early 370, they clearly did not consider it as having ICs of the third-generation. Perhaps it is between second and third generations? Bubba73 ^{You talkin' to me?} 00:27, 5 May 2018 (UTC)

Further discussion

It's not an IC at all. It's a bunch of discrete components bonded after manufacture to a substrate. Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:42, 3 May 2018 (UTC)

But it isn't just a bunch of discrete components. They used photolithography and on-chip metallization and ball bonding, all important parts of monolithic technology. (Though much of non-IBM monolithic circuits use wire bonding instead.) The technologies that they didn't have yet were isolation wells, with reverse biased junctions to keep the transistors electrically apart, and good resistors. They were much closer in reliability to later monolithic circuits than to separately package transistors, and cheaper than monolithic for many years. IBM had automated production and packaging to keep costs down and reliability up, relative to the competition. (All this is in the Pugh book, in case anyone here is interested, chapters 2 and 8.) I agree that they don't go in the monolithic section, but they are enough past discrete packaged transistors, in cost, reliability, and technology, to need some distinction. A separate section in this article would seem fair to me. Gah4 (talk) 18:09, 3 May 2018 (UTC)

• @Gah4:This article is just a list. Why do you think we need a separate section in this article? How about just a sentance in the lede? For example: "Some late transistor computers may have used packaged their discrete transistors in hybrid devices. That along with a footnote on S/360 models should be sufficient. A short section in the main article Transistor computer might be appropriate. If we do this will you also agree to changing this article (making the poll 4:1) Tom94022 (talk) 06:54, 4 May 2018 (UTC)

Why are we making this distinction? (That is an actual question.) It seems to me more politics than science. IBM built with the best technology they could make, in cost and reliability terms. If they hadn't done it, and instead waited for monolithic circuits, they would have lost out to the competition. S/360 processors were already delayed enough. They would have been later, more expensive, and likely less reliable, as was some of the competition at the time. The real advance for S/370 is monolithic MOS technology for main memory, instead of core. They could have built S/370 processors from SLD, but with monolithic DRAM, reasonably comparable to the actual machines. Gah4 (talk) 18:09, 3 May 2018 (UTC)

It's OK to admit you may not have heard of "hybrid integrated circuits", "thick film integrated circuits", and "thin film integrated circuits". They are still around and used where the limitations of monolithic processes are inconvenient. Someone designing a computer with SLT hybrids is in pretty much the same position as a designer using monolithic ICs - the designer is no longer dealing with discrete transistors but instead with a module that provides a logical function, with standardized input and output characteristics. Presumably SLTs also gave significant advantages in size, power consumption, reliability, cost, etc. - and were worth doing before "monolithic" integrated circuits came along with sufficient performance to be useful. That's why they are called "monolithic", to distinguish them from the hybrids. I guess a Pentium Pro is a discrete transistor computer 'cause it's got two dies in one package? --Wtshymanski (talk) 19:08, 3 May 2018 (UTC)

"I guess a Pentium Pro is a discrete transistor computer 'cause it's got two dies in one package?" No, because the dies aren't single transistors.

"Someone designing a computer with SLT hybrids is in pretty much the same position as a designer using monolithic ICs - the designer is no longer dealing with discrete transistors but instead with a module that provides a logical function, with standardized input and output characteristics." Someone designing a computer with DEC Flip Chip modules, even if the modules are made by attaching individual transistors, diodes, etc. with wires coming out of them to the board rather than using flip chip packaging, is also pretty much in the same position as a designer using monolithic ICs in that sense. The same applies to someone designing a computer out of IBM Standard Modular System modules. So the SLT notion of transistors-and-diodes-on-bare-dies attached to the module isn't what makes a difference there.Guy Harris (talk) 20:49, 3 May 2018 (UTC)

The generation before SLT, called SMS, uses three wired transistors on little PC boards. The idea then was that one SMS board would be a few logic circuits, in about the same way as an SLT module or 7400 series IC. Though in the end, they had to make more different SMS modules than they had expected. For the most part, though, designers can treat them as logic blocks. I suspect designers of other three-wire transistor computers used some similar logic module, too. Gah4 (talk) 21:03, 15 May 2018 (UTC)

Sure "wired transistors on little PC boards," SLT modules and 7400 series all perform logic functions but that doesn't change the fact that the transistors in the SLT modules are discrete whereas the transistors in the 7400 series are monolithic. Computers where the primary logic is made from discrete transistors is the subject of this article. The smaller packaging of an SLT module does change the nature of the transistors.Tom94022 (talk) 22:00, 15 May 2018 (UTC)

However, it took the monolithic-IC notion of fabricating the entire circuit on a single die to get us where we are today; unless I'm missing something major, we wouldn't have gotten a machine like the one on which I'm typing this, and wouldn't have gotten something like an IBM z14 mainframe, using technologies such as SLT, even with 55 years of development on it. Guy Harris (talk) 20:49, 3 May 2018 (UTC)

Additional discussion

As it says in the main article a "transistor computer" is synonomous with "second generation computer". as opposed to third generation computer which in the article and with multiple reliable sources is defined as such computers as using "integrated circuits," some going as far as either explicitly stating "monolighic integrated circuits" and or point to the TI or Fairchild invention of such. Furthermore, there are multiple reliable sources that the term "integrated circuit" in its ordinary usage means monolithic integrated circuts. It is a semantics play to say that hybrid circuits such as SLT some makes them integrated circuit in the sense of this article. Tom94022 (talk) 16:37, 16 April 2018 (UTC)

• I think that I was the one that brought up the topic of the 360, and I don't know. (At first I thought the 360 belonged.) I don't know enough about the hardware aspects. My old textbook, Introduction to Electronic Computers, by Gordon B. Davis, second edition, 1971, page 12 says "Third-generation computers becan to be delivered in 1965. The most common system is IBM System/360..." Bubba73 ^{You talkin' to me?} 17:16, 16 April 2018 (UTC)

I would be interested to see how Bell defines second and third generation. I don't have a copy of his book readily available. Again this may be a semantics issue. Tom94022 (talk) 18:29, 16 April 2018 (UTC)

If by "his book" you mean Computer Structures: Readings and Examples, there's an online version, and, on page 39 in part 1 "The structure of computers", chapter 3 "The computer space", it says:

The generations are best defined solely in terms of logic technology: The first generation is that of vacuum tubes (1945 ~ 1958), the second generation is that of transistors (1958 ~ 1966), and the third generation is that of integrated circuits (1966~). In fact, current usage describes hybrid logic technology machines, such as the IBM System/360, as third generation, and so this extension must he included. What will be called fourth generation is yet to emerge; most likely it will he medium and large scale integrated circuits with possibly integrated circuit primary memory.

Table 1, "The computer-space dimensions", on page 40, has a section "Logic technology", which lists, sorted in order by increasing speed (rather than date of introduction), "Mechanical", "Electromechanical", "(Fluidics)", "Vacuum tube", "Transistor", "Hybrid", "Integrated/IC", and "Medium to large-scale integrated/MSI ~ LSI", separating "Hybrid" from "Integrated/IC", and labeling "Vacuum tube" as the first generation, "Transistor" as the second generation, and "Integrated/IC" as the third generation, leaving the "Generation" column blank for "Hybrid".

The book discusses SLT on page 564, in a section on S/360, saying "The logic of the 360 series is realized in a hybrid technology, composed partly of integrated-circuit techniques and partly of the solid-state techniques standard in second-generation machines.", and that

Although SLT differs fundamentally from integrated-circuit technology, the overall size of the final printed-circuit boards is about the same. At the time the decision was made to develop the technology, it was unclear that integrated-circuit technology would reach mass-production state. Thus the SLT program was an intermediate design prior to integrated-circuit technology. The two approaches are about the same from the standpoint of reliability, especially when one considers the soldered printed-circuit mounting. The number of connections to the printed-circuit board are about the same.

So it sounds as if Bell and Newell 1) consider hybrid technologies such as SLT "fundamentally" different from (monolithic) IC technology but 2) may not consider them to be so different as to put them in separate generations. Guy Harris (talk) 19:17, 16 April 2018 (UTC)

Thanks for the link. Further on in Bell they state:

The logic of the 360 series is realized in a hybrid technology, composed partly of integrated-circuit techniques and partly of the solid-state techniques standard in second-generation machines
...
The semiconductor elements (diodes and transistors) are produced independently, using standard semiconductor production techniques on a wafer. The semiconductors are then cut and bonded to the substrate, and the complete SLT logic unit is encapsulated

I think this makes it pretty clear that the chips used in SLT are discrete components. Encapsulating these chips on an SLT substrate does not make it into an IC as the term is currently defined any more than encapsulating transistors on a pcb makes the assembly into an IC (which was done on militarized second generation computers). Bell does state that "current usage," that is 1971 usage, describes hybrid machines as third generation, but 2018 usage of "integrated circuit" means monolithic which relegates most S/360 to second generation.Tom94022 (talk) 16:19, 17 April 2018 (UTC)

A hybrid integrated circuit is not a discrete transistor. A discrete transistor has a package with 3 wires leading to the base, emitter and collector regions of the transistor within (or equivalent in FET - were any transistorized computers made with discrete FETs?). An SLT or hybrid IC has many internal devices, but the package leads no longer connect one-to-one with internal regions of a single transistor. Hybrid ICs were an innovation over discrete transistors. The individual "chip" transistors in a hybrid IC are of no more worth than the individual diffused transistors in an integrated circuit - the design is based on the packaged device. --Wtshymanski (talk) 20:00, 16 April 2018 (UTC)

From the way SLT is built, I suspect that the failure rate is more like the failure rate of ICs, than of discretely packaged (three wire) transistors. But it seems to me that the things that really makes third generation computers is monolithic IC main memory. Though as I understand it, the first use of monolithic IC memory in the memory subsystem of a commercial machine is the storage keys in the 360/91, built from 16 bit chips. But otherwise, the main memory of the 360/91 is 750ns core. The 360/85 introduced cache, I believe made from ICs. S/370 uses ICs and monolithic memory. Gah4 (talk) 20:20, 16 April 2018 (UTC)

As far as I know, nobody has claimed that hybrid integrated circuit is a discrete transistor. (If somebody believes that I said that, they've significantly misread what I've said.) The question is whether the "silicon planar glass-encapsulated transistors and diodes" (to quote IBM's paper on SLT) on a hybrid integrated circuit counts as discrete transistors and diodes. Packaging the transistors by encapsulating them in glass and directly soldering the "individual chip devices" together is definitely an innovation, but I, at least, see fabricating multiple transistors on one die, with the connections between them being part of the fabrication process rather than being done after the fabrication process and cutting the wafters into dies, as being a much more significant innovation.

So, are hybrid ICs significant enough that systems using hybrid ICs don't belong in this page? Or are they different, but not different enough, so that this page should perhaps be expanded to, in effect, a list of post-vacuum-tube and pre-monolithic-IC computers, perhaps specially noting the hybrid-IC computers? Guy Harris (talk) 21:23, 16 April 2018 (UTC)

We should really have a WP:RS one way or the other, but it looks like there might not be one. If some sources say 2nd, and some say 3rd, it might be that we just can't say. As well as I know it, IBM considered IC technology, but didn't believe that it was ready on the scale that they needed it. Gah4 (talk) 23:20, 16 April 2018 (UTC)

Well, using that argument, no printed circuit board with discrete transistors on can be considered to contain discrete transistors since the three connections to each transistors are rarely brought out to the board's external connections. That is a nonsensical argument.

'Integrated circuit' is a rather nebulous term. Early integrated circuits were just substrates with small form-factor discrete components on them some even covered in some sort of resin. I would regard such circuits as containing discrete transistors in just the same way as a printed circuit board does, because they are capable of being removed and reused (if with difficulty, but it could be done). Only if the transistors are fabricated as part of the integrated circuit do they cease to be discrete because they cannot be individually removed.

Arguably: these are also integrated circuits and do not contain discrete vacuum tubes because they cannot be removed and used individually. 86.153.129.239 (talk) 13:12, 17 April 2018 (UTC)

I believe that IBM well documented the decision to use SLT as we know it. The question here is the naming of generations. SLT is a technological improvement of SMS (transistors and other components soldered onto PC boards), but not all the way to monolithic technology. TTL was available at about the time of S/360, but it seems not quite ready enough. There were enough delays in getting S/360 out, and going to monolithic circuits might have added to the delay, or at least it was a reasonable expectation at the time. Reliability was, and is, very important. Maybe SLT should be 3rd generation, and monolithic ICs 4th generation, but that isn't the way the numbering went. As far as I know, you can't get the individual transistors off an SLT ceramic substrate. Gah4 (talk) 14:06, 17 April 2018 (UTC)

As I remember the technology, one starts with a ceramic substrate and applies metallised path ways to act as connections and resistors. Discrete transistors (and very likely diodes) are then added to the substrate. They are then encapsulated. Once the encapsulation is removed, the transistors would be easily recoverable. The transistors were certainly discrete parts being manufactured entirely separately from the ceramic substrate. 86.153.129.239 (talk) 14:45, 17 April 2018 (UTC)

What is an integrated circuit

In general usage integrated circuit has come to refer to the single-piece circuit construction originally known as a monolithic integrated circuit.

The above is quoted from the Integrated circuit article and has multiple RS's beyond the two. Maybe someone some place called SLT an "IC" but in today's terminology they are not. So in today's terminology the early S/360's are transistor computers. We can footnote the fact that they used hybrid assemblies but they should not be excluded from this list Tom94022 (talk) 16:57, 17 April 2018 (UTC)

But is the actual question 3rd generation or not? Today, and in hindsight, we define IC in a specific way. This actual article says until hybrid and monolithic integrated circuits displaced them, so seems to exclude hybrid circuits. Reading IBM's 360 and Early 370 Systems there is a whole chapter on the development of technology, starting from the alloy junction transistor. It seems that just about all the developments needed, planar technology and glass encapsulation, were used in SLT. The technology of SLT is way past that used in previous generations of transistorized computers. DCTL, not used in SLT and S/360, has the multiple emitter transistors that went into TTL, but could have been used. Gah4 (talk) 22:37, 17 April 2018 (UTC)

"Just about all the developments needed" for what? All the developments needed for monolithic ICs? Were any transistors-in-cans also manufactured with a planar process? (My copy of Pugh et al is in a box somewhere, and there's no electronic version I could find.) Guy Harris (talk) 23:26, 17 April 2018 (UTC)

Everything is planar now, cans or not. Before planar there were alloy junction transistors, with dopants diffused in from both sides of a piece of germanium. The mesa transistor is etched down. I suppose using Pugh as a primary reference isn't the best choice. I didn't realize before that SiO2 masking was used on germanium before silicon. Using ball bonding to attach the chip onto the substrate was pretty advanced. It seems that the main reason for not choosing monolithic was yield, figuring that if each transistor had a 50% chance of working, two would have 25% chance, and three 12.5%. They didn't yet understand that they are not statistically independent. But also, they needed a process that would scale up well, with no surprises. Gah4 (talk) 00:13, 18 April 2018 (UTC)

This is off-topic for the talk page for this article, which is about transistorized computers. We don't need to have the theological debate here and perhaps not anywhere. --Wtshymanski (talk) 17:26, 19 April 2018 (UTC)

Off topic for an article. Also, I don't see that I mentioned god. Gah4 (talk) 18:48, 19 April 2018 (UTC)

The machine on which I'm typing this is "transistorized", in that it uses transistors rather than vacuum tubes, but there are several billion of them fabricated on the die containing the CPU. List of transistorized computers doesn't include all computers "transistorized" in that sense; the question is where the line should be drawn - should it include, or exclude, hybrid circuits? - so, at least to that extent, the question of whether hybrid circuits are 1) enough like circuits made of individually-packaged transistors so that it should include hybrid circuits, 2) enough like monolithic integrated circuits so that it should exclude hybrid circuits, or 3) different enough from both of them that it should, probably, exclude hybrid circuits is relevant. Guy Harris (talk) 18:16, 19 April 2018 (UTC)

While the 2nd, 3rd, 4th generation system doesn't seem to allow for it, given that this page is title transistorized, it seems to me that we can have a page for hybrid circuits like SLT. That they are enough of an advance over previous technology, though not yet the monolithic technology of S/370. Which other machines have similar technology, that could be included on such a page? Gah4 (talk) 18:48, 19 April 2018 (UTC)

I don't think we need a new page as long as we accept that in current terms IC means monolithic. IBM apparently never described SLT as a hybrid IC and in the two contemperaneous RS IBM didn't even describe SLT with either the term hybrid or integrated. The current Wiki pages for Transistor computer, List of transistorized computers and Integrated circuit are also now consistent with the modern meaning of IC. So it seems like almost POV that the packaging innovations of SLT some makes computers built with it not transistor computers. I suggest we should add them to the list and note that "SLT used a microelectronic module to package discrete transistors and diodes within the modules then mounted to circuit boards." "Microelectronic" and "module" are IBM words or we can use words to that effect. Tom94022 (talk) 20:04, 19 April 2018 (UTC)

But they still aren't discrete transistors. Have a look at W. Grieg Integrated Circuit Packaging, Assembly and Interconnetion, page 63 [4]. How is the hybrid integrated circuit described here different from an IBM SLT? Just because IBM reused to use the insdustry term for their device? --Wtshymanski (talk) 02:13, 21 April 2018 (UTC)

Thanks for the reference, Gregg proves the point since he clearly distinguishes between hybrid circuits and IC's

5.2 - The Hybrid Circuit
A hybrid circuit is definded as as assembly containing both active semiconductor devices (packaged and unpackaged) ...

Figure 5-2. Hybrid Circuit with Transistors and Diodes (Circa 1960s)

5.2.1 — The IC and the Hybrid Circuit
With the emergence of the IC in the late ’60s, it was widely believed that the hybrid circuit would quickly lose significance as a packaging technology. However, when the IC became readily available it was quickly incorporated into the hybrid (Figure 5-3) as the active device replacing the earlier diodes and transistors. ...

Integrated Circuit Packaging, 2007, p63-4

The figure u reference is indeed just like an SLT module but Gregg clearly distinguihes it from an IC as we should in this article. Tom94022 (talk) 06:50, 21 April 2018 (UTC)

@Gah4: there doesn't appear to be any support for a separate "list of hybrid circuit computers" and such a list might be original research. Do you have a position on whether we can add early S/360 computers to this list? If so please, add your position to the poll. Since this article is a just a list you might take into consideration that its main article notes that some early transistor computers still used tubes so that it seens reasonable to include a note therein that IBM's early System/360 computers packaged their discrete transistors in hybrid circuits (IBM SLT circuits) and then list them herein, again with an appropriate note. This avoids the original research of defining a hybrid 2.5^th generation.

Taking a position will make it easier for an adminstrator to decide whether we have achieve a rough consensus or not. Tom94022 (talk) 18:30, 15 May 2018 (UTC)

References

Honeywell?

The only Honeywell machine listed is the H-200. What about the H-800 and the DDP line used for the IMP and CHIMP on the ARPAnet? Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:44, 16 April 2018 (UTC)

Added Honeywell 800 (6000 transistors, 30,000 diodes in the processor). The Honeywell 316 article says it used DTL ICs. --Wtshymanski (talk) 02:01, 21 April 2018 (UTC)

Computer Control Company's DDP-116, of which the Honeywell 316 and other members of the Honeywell 16 series are descendants, and DDP-24 were probably discrete-transistor computers. Guy Harris (talk) 02:19, 21 April 2018 (UTC)

Was the H-8200 transistor or IC? Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:26, 23 April 2018 (UTC)

Page 83 of the November 1, 1965 issue of Electronics magazine says that it will use ICs, and this page on 1970's computers says "My first real programming training was a Cobol course on a Honeywell 8200 which was among the first generation of commercial computers built using integrated circuits, rather than individual transistors." Guy Harris (talk) 18:11, 23 April 2018 (UTC)

For what it's worth

I don't understand all of the electronic terms, but IBM said that the SLT is not an integrated circuit. It is a Hybrid Integrated Circuit or Hybrid Circuit. From IBM's 360 and Early 370 Systems, by E. Pugh, et. al., p. 106. Bubba73 ^{You talkin' to me?} 19:38, 3 May 2018 (UTC)

Yes, but what do we do here? Should we make the page List of hybrid circuit computers? Are there enough for such a page? From Pugh, IBM figured that SLT is about 100 times more reliable than discrete (individually packaged) transistors, and about 10 times less than monolithic circuits would be. On the logarithmic Moore's law scale, that makes then 2/3 of the way, which agrees with what looks like they are 2/3 of the way in technology. IBM outsold competitors that did use monolithic circuits, which along with other reasons showed that SLT was the right technology for the time, at the scale that was needed. Gah4 (talk) 12:30, 4 May 2018 (UTC)

In answer to @Gah4:'s question, neither a parent article "Hybrid circuit computers" nor a subordinate article "List of hybrid circuit computers" makes much sense since there is likely not any reliable source for such articles whereas there are many reliable sources for this article and its parent, Transistor computer, as the second generation of computers. We have reliable sources that the transistors in the SLT hybrid circuits were discrete transitors which according to the definition makes computers using them transistor computers. Yes the packaging into SLT hybrid circuits improved certain characteristics but that doesn't change the discrete nature of the transistors. So, @Bubba73: what's wrong with listing them herein with the appropriate annotation that the discrete transistors were packaged in hybrid circuits? Tom94022 (talk) 18:18, 4 May 2018 (UTC)

I'm not saying that there is anything wrong with including them. I don't know much about the electronics technology, so I'm not going one way or the other. I did bring up the question of whether or not the 360 should be here. I used a 360/65 back in the day, but I've never touched one and I am not knowledgeable about electronics. Bubba73 ^{You talkin' to me?} 23:14, 4 May 2018 (UTC)

OK, thanks for asking the question. May I suggest u post neutral in the Poll section above so that if and when we have to go to an administrator yr position is clearly summarized? Tom94022 (talk) 23:59, 4 May 2018 (UTC)

@Tom94022: Here's an example of how it could be done: Draft:List_of_integrated_circuit_computers. That leaves a question of duplicates between transistor/IC list - some computers may end on both. Is a hybrid computer more transistorized, or more IC?

Or it could be done the same way (ignored on list) as for earlier hybrids (or near hybrids) before transistor: Zuse Z2, IBM SSEC, SEAC (computer), SWAC (computer)?, ERA 1101, Harvard Mark III, ...--89.25.210.104 (talk) 12:34, 18 May 2018 (UTC)

There is no list of third generation (monolithic IC) computers but if there were it would have different computers than you have posted in your draft article. I suggest your draft article is more accurately entitled "List of hybrid circuit computers" since it mixes computers with hybrid circuits using ICs (Martac 420), with hybrid circuits using discrete transistors (IBM) and with hybrid circuits using diodes (DEC). Those computers on your list using monolithic ICs in hybrid packages or not should be considered for addition to History_of_computing_hardware_(1960s–present)#Third_generation or creating an article that accurately is a "List of third generation (monolithic IC) computers". Those computers with hybrid circuits using discrete transistors belong in this list and I thank you for finding some more. Tom94022 (talk) 18:21, 18 May 2018 (UTC)

Another old textbook

Computers: Appreciation, Application, Implications - an Introduction, by Adams and Haden, 1973, p. 231-232, "Third-generation computers are identified by their integrated circuits ... The major transition to third-generation computers was in 1965, when IBM began deliveries of its third-generation System/360." Bubba73 ^{You talkin' to me?} 00:35, 5 May 2018 (UTC)

Why are we making this distinction? (That is an actual question.)

As I asked in earlier: Why are we making this distinction? (That is an actual question.). That is, why do we care about the difference? Not that we shouldn't, but the actual reason might determine how the article should be written. Often enough, is seems to me, distinctions about naming are made for political reasons, and not engineering/technical reasons. There was much discussion inside IBM at the time, but later they determined that SLT was cheaper and more reliable than monolithics throughout the SLT era. They had some competition at the time, using monolithic circuits, with names now mostly forgotten as computer companies. In reliability, SLT is 2/3 of the way on the logarithmic Moore's law scale between individually packaged transistors and monolithic logic. Gah4 (talk) 22:16, 15 May 2018 (UTC)

Distinction between three-wire transistors and hybrid circuit technologies such as SLT, or distinction between hybrid circuit technologies and monolithic IC technologies?

Presumably the former distinction would be made for the reliability reasons you mention.

The latter distinction is made because hybrid circuits wouldn't have given me the 64-bit superscalar 4-way (chip) multiprocessor 2-way threaded processor in the computer that's sitting on my lap as I'm typing this. In circuit density, SLT is a long way from where monolithic logic is today, and probably wouldn't have made it much further, unlike monolithic logic.

So maybe we treat S/360 differently from the three-wire-transistor-based machines, but I'd disagree with treating it similarly to monolithic-integrated-circuit machines like S/370, much less like my MacBook Pro - or a z14. Guy Harris (talk) 00:35, 16 May 2018 (UTC)

OK, but it isn't fair to compare SLT to today's monolithic. It seems that for S/370, IBM put the monolithic chips on the same ceramic substrates as SLT. They might have had a little more logic than SLT, but not a lot more, so the density wouldn't be much better. As I wrote a few times, it was the 1K bit DRAM chips that really made S/370 different. They knew when designing SLT that monolithic was on the way, it just wasn't ready yet. Gah4 (talk) 05:11, 16 May 2018 (UTC)

Reminds me, there used to be the distinction SSI, MSI, LSI, VLSI, for different density of monolithic circuits, but that stopped at VLSI. (I used to know people trying to extend it, such as RHSI, that is, Really Humongous Scale Integration.) There are plenty of technology changes made of the years between SSI TTL and your MacBook, much more than between SLT and SSI TTL. Gah4 (talk) 05:27, 16 May 2018 (UTC)

For this article there is no distinction of substance between three wire transistors and three solder bump transistors - they are both discrete transistors. There are multiple reliable sources for an integer system of classifying computers by types of logic circuits - four generations, tube, discrete transistor, monolithic IC and microprocessor. Distinctions like reliability and design style are strawman arguments - irrelevant to this taxonomy and this discussion. Given the taxonomy we can classify each of the early S/360's into only one of these four categories or we can ignore the most successful computers of the 1960s. Tom94022 (talk) 07:09, 16 May 2018 (UTC)

"There are plenty of technology changes made of the years between SSI TTL and your MacBook, much more than between SLT and SSI TTL." The change between SLT and SSI TTL is "we can make a circuit, not just a single transistor or a pair of diodes, using fabrication technologies"; that's the underlying idea that's used in the Core i7 in my MBP. Yes, they went from TTL to CMOS, and they used different forms of photolithography, but the concept of putting a circuit on a die, rather than a single transistor or pair of unconnected diodes, strikes me as the key difference between circuits made of separately-fabricated transistors and monolithic ICs, so, no, I'm unconvinced that the difference between SSI TTL and my MacBook is fundamentally bigger than the difference between SLT and SSI TTL. I view the difference between SLT and monolithic ICs of any sort as a conceptual leap, and the differences between various forms of monolithic IC as smaller steps.

Perhaps to somebody involved in the physics of semiconductor fabrication, those steps are significant, but from the point of view of somebody at the circuit level, they're manufacturing refinements following the conceptual leap to "not just making a transistor in the fab, but making a gate, or collection of gates, or set of connected gates, or..." in the fab. Guy Harris (talk) 07:59, 16 May 2018 (UTC)

They are still significant at the circuit level, at some of the points along the way, but that isn't the question I am trying to ask. For comparison, some years ago there was much discussion over the boxes used to connect to DSL and cable internet, specifically claims that they weren't modems. I suspect that the reason is the connection between the word modem and dial-up technology. There were claims such as the signal being digital on both sides (no), or some such. The boxes do modulation and demodulation, and so are modems, even if it is the same term used for dial-up boxes. But some terms have superstitious meanings to some people. Gah4 (talk) 18:01, 16 May 2018 (UTC)

OK, one example which I happen to remember. About at the change from 0.8 micron to 0.6 micron, or early 1990s, you have to consider wires as distributed RC circuits. Previously, you consider them lumped capacitors driven by the output current of the source. It isn't fabrication differences, but scaling laws. The change from bipolar to MOS is pretty significant at the fabrication level, and also at the circuit level. The change from metal gate to self-aligned silicon gate was pretty important, too. There are many more that I am not thinking of right now. Gah4 (talk) 18:01, 16 May 2018 (UTC)

So what is the question you are trying to ask? If it's "what superstitious meaning is being applied to "intergrated circuit"?", the answer, at least for me, is "none". Guy Harris (talk) 18:23, 16 May 2018 (UTC)

This discussion is becoming repetitive, at least in part irrelevent and somewhat uninteresting. The only "actual question" remaining at this point is do you, Gah4, have any opinion as to the addition of some early S/360 systems to this list? Please take a postion, for, neutral or against. Tom94022 (talk) 18:55, 16 May 2018 (UTC)

OK, what I actually mean by superstitious is marketing sloganeering. From Pugh, IBM was worried about competition using monolithic circuits having a marketing advantage. About eight months after S/360 was announced, RCA and Scientific Data Systems announced computers built with monolithic circuits. The fact that those names are not well known today as computer companies should tell you something. In marketing terms, SLT is discrete transistors, and IBM knew that at the time. But in technical terms, SLT was cheaper and more reliable than monolithic circuits through most of its life. (Pugh says half the price through 1968. I am not sure of the reliability crossover.) I suppose I would most like to see them as a separate section on this page. Gah4 (talk) 01:07, 17 May 2018 (UTC)

@Gah4: since we agree that "SLT is discrete transistors, and IBM knew that at the time" would u mind adding to the poll that u are in favor of adding early S/360 to this list? With the poll at four in favor, one neutral and one against I think we have rough consensus and I can ask an administrator to close the discussion. Tom94022 (talk) 19:25, 17 May 2018 (UTC)

But SLTs aren't discrete transistors. It would be clearly wrong to add the early System 360s to this list. --Wtshymanski (talk) 20:22, 17 May 2018 (UTC)

No one ever said SLTs are discrete transistors but most editors and RS's agree they use discrete transistors just like pcbs use discrete transistors. Actually polling can indicate rough consensus - your POV is not sufficent to exclude early S/360s from this list. Tom94022 (talk) 21:55, 17 May 2018 (UTC)

Transistors in an SLT are not discrete transistors. They have no packaging aside from the SLT package, their emitter/base/collector terminals are not accessible to the outside world. A transistor on a thick film integrated circuit is not a discrete transistor. --Wtshymanski (talk) 00:39, 18 May 2018 (UTC)

It seems to me that in marketing terms, that they are discrete transistors. IBM knew this, and started using hybrid integrated circuit in discussions of the technology. But note that they are not individually packaged, which is normally also part of discrete device specifications, especially in the technical sense. The reason for this section, is that I want to know if it is marketing sense, or technical sense, that is important here. Gah4 (talk) 00:55, 18 May 2018 (UTC)

Transistors in an SLT module are definitely not like transistors in a monolithic IC - their connections to other transistors are all on the substrate, not on the die, unlike monolithic ICs where some connections are on the die itself.

So, if we're going to draw a distinction between transistors in cans with three wires and transistors on an SLT substrate, and deem only the former to be discrete transistors, we should 1) update Transistor#Packaging to remove the paragraph saying

Discrete transistors as transistor chips may also be packaged in hybrid circuit devices. The IBM SLT module of the 1960s is one example of such a hybrid circuit module using discrete glass passivated transistors (and diodes). Other packaging techniques for discrete transistors as chips include Direct Chip Attach (DCA) and Chip On Board (COB).

2) perhaps update transistor computer to say

These machines remained the mainstream design into the late 1960s, when hybrid and monolithic integrated circuits started appearing and led to the third generation machines.

and 3) modify the list itself to say

Discrete transistors were a feature of logic design for computers from about 1960, when reliable transistors became economically available, until hybrid and monolithic integrated circuits displaced them in the mid 1960s.

so as to clearly indicate that we're separating discrete transistors from transistors in hybrid ICs and that we're also treating hybrid ICs and monolithic ICs differently. Guy Harris (talk) 04:35, 18 May 2018 (UTC)

Excellent idea. I have followed up on your suggestion. --Wtshymanski (talk) 04:06, 19 May 2018 (UTC)

Oops! It appears that integrated circuit is a page about monolithic integrated circuits, and there's a separate hybrid integrated circuit page, so if we're going to say that this page should discuss computers not made with either type of IC, we should say

Discrete transistors were a feature of logic design for computers from about 1960, when reliable transistors became economically available, until hybrid and monolithic integrated circuits displaced them in the mid 1960s.

Sorry about that. Guy Harris (talk) 01:19, 20 May 2018 (UTC)

Regarding a separate section, this article is just a chronological list so a separate section doesn't make much sense. I suggest we can add a separate section to the main article and footnote the added System/360s in this article. Tom94022 (talk) 19:25, 17 May 2018 (UTC)

It tells me that computing with Big Blue was hard. They were big even before S/360 came out using SLT and Spectra 70 came out using monolithic ICs, so I'm unconvinced that RCA's departure from the computer business tells me that SLT vs. monolithic ICs had all that much to do with it, if anything. SDS was only in a specialized segment of the market, so it's not clearer that their failure had much to do with their choice to use monolithic ICs, either.

I see SLT as a stopgap technology, chosen because IBM (perhaps wisely) chose to be conservative and not immediately go with monolithic ICs. In a world of shrinking transistors, the future lies with making circuits in your fab, with several of those smaller transistors on a die (with "several" eventually becoming "billions"), rather than making smaller and smaller dice and mechanically placing them on a substrate. Guy Harris (talk) 01:37, 17 May 2018 (UTC)

OK, so one is that IBM automated SLT production, much more than others had automated their processing. Though it seems that IBM didn't patent some of the automation technology that they could have, so others started using it pretty soon. ASLT production was contracted out to TI, as IBM was too busy making SLT. IBM knew how to make monolithic circuits, but they didn't yet know how to make them cheap and reliable. Resistors had a 20% tolerance. I am not sure when it was known, but an important idea in monolithic circuit production is that resistors have a large tolerance, but ones on the same chip tend to be close. Circuits are designed to take advantage of that. Having different circuits work reliably together is very important in large systems. Margins can shift with temperature and voltage, making system unreliable. SLT uses the planar process used for monolithic circuits, so working with it would also help them in production of monolithic circuits. That is a big improvement over the Alloy-junction_transistor used in earlier processors. I wonder, do we know which processors were produced with planar technology? Gah4 (talk) 02:53, 17 May 2018 (UTC)

The story I remember hearing for about 40 years, is that it was the space race, and demand by NASA, that got the economy of scale of monolithic circuits started. Early on, they were expensive, and the only way to get the price down was to make enough of each one. But, being expensive, there wasn't the demand. Pugh says that SLT was half the price of monolithic until 1968. If RCA and SDS were building computers with more expensive parts, the prices would have been higher, too. The way we thing of IBM today, as the big computer company, is very much based on the success of S/360. Now, the price and reliability of SLT and competing monolithic circuits wasn't known to IBM from the start. At some point, it was the computer companies' demand for TTL that got the economy of scale up, and the price down, which then led to cheaper, faster, computers, and also affordable TTL chips for hobbyists. But yes, that isn't conclusively the reason RCA and SDS are out of the computer market today. Gah4 (talk) 00:20, 18 May 2018 (UTC)

"But yes, that isn't conclusively the reason RCA and SDS are out of the computer market today." Probably not even the most important reason. Part I of this exhibit from US v. IBM says, among other things:

RCA had an early start in the computer business, but delivered only nine computer systems commercially prior to 1960.

and, in general, says RCA's computer business wasn't in great shape prior to the introduction of Spectra 70.

Part II of the exhibit says:

Within IBM the announcement of Spectra was noted in a memorandum from C. E. Frizzell, President of GPD, to T. J. Watson, Jr. Frizzell wrote that the series offered better price/performance than IBM in CPU-memQry speed, magnetic tapes and high speed printing, but assured Watson that he was "moving rapidly to meet this challenge and expect to respond effectively in the very near future".

so price/performance might not have suffered from the choice of monolithic IC technology. It appears, from that document, that IBM reduced additional shift charges in response to various competitors machines, suggesting that not all of the price/performance issue was a (production) cost/performance issue, however.

The document also says that "Despite the numerous problems experienced by the Spectra 70 series, during the period of its life (1965-1969) RCA enjoyed considerable success with its computer business."; their ultimate departure from the computer business seems to have been a business issue rather than a technology issue. (Unlike IBM, RCA not only also made TVs, they made TV dinners starting in 1970.)

(That exhibit is pretty interesting as a historical record. For example, it quotes Fred Brooks as holding the opinion that dynamic relocation hardware "was unnecessary for time-sharing or any other purpose". I'll have to try to find the reference, to see if it indicates what led him to that conclusion at that time; given that it's Fred Brooks, it probably was a non-unreasonable conclusion, but presumably he either 1) meant it was unnecessary at that time or 2) made assumptions about the future that turned out not to be true.) Guy Harris (talk) 02:25, 18 May 2018 (UTC)

About 20 years ago, I was wondering about the future of hobbyist digital electronics, when TTL would fade away. It seems to be still around, but also affordable FPGAs and FPGA software. (20 years go, FPGA software licensing was too expensive for hobby use.) Gah4 (talk) 00:20, 18 May 2018 (UTC)

some transistors are more equal than others

To widen the discussion somewhat, this article is on discrete transistor computers, between vacuum tube computers, and those from monolithic circuits. By the time of vacuum time computers, vacuum tube technology was reasonably well developed, but that isn't true for transistors. After the original point contact transistor, which as far as I know didn't go into any computers, there is the grown-junction transistor then the alloy-junction transistor and finally the planar process. Each stage improved reliability and reduced cost. Grown-junction and alloy-junction were most commonly used with germanium, and the planar process with silicon. If the technology is known for enough computers, it would seem useful to separate them that way, and by year within each technology. Much of the advantage of SLT is due to using planar process transistors, in addition to improvements in the process itself. Gah4 (talk) 05:16, 18 May 2018 (UTC)

True but irrelevent to this discussion. The the first planar transistor, the 2N1613, was introduced April 1960. The technology was licensed throughout the industry. So the use of planar transistors in SLT in 1964 was nothing new and planar transistors were likely used in most if not all transistor computers introduced after 1960. Tom94022 (talk) 21:10, 18 May 2018 (UTC)

IC vs Monolithic IC in lede

There is simply no dispute that in today's usage IC means monolithic IC, see e.g. the lede to the article on integrated circuits and many RS including but not limited to the Bell book. By leaving the qualification out of this lede and simply stating the transistor computers were succeeded by integrated circuits we create an ambiguity since the term "integrated circuit" has been used to describe many other types of devices other than monolithic ICs implying that computers built with such devices succeeded transistor computers. There is no RS for such an implication only POV. Tom94022 (talk) 22:19, 21 May 2018 (UTC)

What was described as an "integrated circuit" other than a monolithic integrated circuit? I've certainly never seen such a usage for, e.g., a PCB. Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:49, 22 May 2018 (UTC)

Hybrid integrated circuits. For example, the IBM paper "Solid Logic Technology: Versatile, High-Performance Microelectronics" says

A new microelectronics technique called Solid Logic Technology, or SLT, is utilized in the new family of IBM/360 computers. This new technology provides a hybrid, integrated circuit module which combines discrete, glass-encapsulated silicon transistors and diodes with stencil-screened land patterns and precision passive components.

A Google search for "hybrid integrated circuit" shows other use of that phrase. Guy Harris (talk) 19:02, 22 May 2018 (UTC)

Thanks, this is perfect example as to why we need the qualification in the lede. Without it some may interpret 3^rd generation computers as including those built with hybrid devices whereas it is clear from the many sources (e.g. Bell and also Blaauw) that they are referrring to monolithic ICs as the logic circuitry defining 3^rd generation computers. BTW there were hybrid devices incorporating monolithic ICs which would be 3^rd generation. Tom94022 (talk) 16:52, 23 May 2018 (UTC)

This is an example from the Drummer 1963 reference of usage of IC in other than monolithic:

For the purpose of this discussion microelectronics is divided into five different systems, namely:-

(a) Standard size micro-components (micromodules)

(b) Thin film integrated circuit* (microcircuit)

(c) Semiconductor integrated circuit* (solid circuit)

(d) Multi-chip semiconductor circuit (multi-chip circuit)

(e) Hybrid integrated circuit* (hybrid circuit)

The terms in brackets are shortened colloquial titles and will be used here for convenience. Examples of the first four constructions are available commercially.

Granville in Drummer, 1963, p.310

In today's terminology only item (c) is an integrated circuit. Tom94022 (talk) 16:57, 23 May 2018 (UTC)

Note the use of discrete in the above quote from the IBM paper. Also note the use of discrete in "Steps in manufacturing Solid Logic Technology hybrid wafers" . There is simply no question that the individual transistors used in the SLT hybrid devices are accurately described as discrete transistors. Tom94022 (talk) 16:52, 23 May 2018 (UTC)

The quote from Drummer doesn't mention the unqualified term "integrated circuit". Do you have a link to the entire paper? Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:19, 24 May 2018 (UTC)

Sorry I don't have a link to the entire paper but I have poked the Google version (searched within the Google book on "integrated circuit") enough to conclude that Drummer's 1963 paper does not have an unqualified definition of "integrated circuit." Note therein Granville defines his term "solid circuit" for what is today termed an IC (monolithic). If u look at the use of IC in both Bell and Blauuw it is clear they meant monolithic IC in defining their third generation computers and most modern RS's shorten monolithic IC to just IC. So adding "monolithic" to the lede of this article avoids potential confusion with or mis-application of obsolescent definitions of qualified ICs as in Drummer. Tom94022 (talk) 18:36, 24 May 2018 (UTC)

Tom and I have been discussing this on my talk page. The choice of the word "discrete" in the image caption was mine. The Computer History Museum's description was more nuanced, referring to "separate transistors." Regardless, this is a list article and it states its inclusion criteria clearly, transistorized but not monolithic integrated circuits. It is not discussing computer generations. So the System 360 belongs here but should include a note that it used hybrid integrated circuits, as should the already-listed IBM 1130, which used the same SLT technology as the 360.--agr (talk) 03:02, 25 May 2018 (UTC)

Arnold, note the parent article, Transistor computer, categorizes these computers as "second generation computers." Does that in any way change yr recommendation? FWIW I was thinking of adding a section on hybrid packaging (Computers using discrete transistors in hybrid devices) to the main article and then listing the models here. Tom94022 (talk) 17:51, 26 May 2018 (UTC)

No change in my recommendation. The parent article, Transistor computer, needs a lot of work. It doesn't even mention the IBM 7000 and 1400 series, which dominated the IT industry, nor the early DEC machines. The generational transitions were not as clean as some some authors would like. I think the best thing, as you suggest, is to add a section on the 360 and hybrid packaging as a bridge between second and third generation. It would be worth mentioning that IBM achieved circuit densities in their hybrid packages equivalent to early monolithic ICs. The chips used in the Apollo Guidance Computer, the first successful IC computer, had 6 transistors each (3 in the Block I design). In all other ways the 360 represented a clear cut generational change. As long as we tell that story clearly, it doesn't matter where we list the 360.--agr (talk) 03:23, 27 May 2018 (UTC)

As I mentioned earlier, the transitions from grown-junction transistor to alloy-junction transistor and finally the planar process were very significant, yet are ignored in much of the discussion. Gah4 (talk) 08:44, 27 May 2018 (UTC)

A section on the evolution of transistor technology and its impact would be welcome. I've added material on IBM and the 360.--agr (talk) 12:11, 27 May 2018 (UTC)

I suggest such a section in the main article is overkill, perhaps all we need a statement performance improvements were in part driven by the evolution of transistor types along with a link to Bipolar transistors - note this section links to separate articles on most (all?) types of bipolar transistors including the three Gah4 cites. In any event, the absence of such a section or link is not a valid reason for not adding S/360 to this list - after all their logic was primarily discrete bipolar planar transistor chips. Tom94022 (talk) 19:18, 27 May 2018 (UTC)

The Bipolar transistors section you suggest is a laundry list with no context. If there is sourced material saying which transistor improvements impacted computer development it would be very relevant. But this discussion belong in the parent article's talk page. Is there any objection still to including the 360 on this list with a note about hybrid integrated circuits? The IBM ref I added to the parent article explains what happened and why (monolithic technology wasn't mature enough to bet the company on) in great detail. Indeed the very name "Solid Logic Technology" was IBM marketing's clever way to finesse the issue.--agr (talk) 14:26, 28 May 2018 (UTC)

I agree a section on the impact of transistor technology on early computers would be a useful addition to the main article but have looked for an RS without any success so far. Perhaps all we can do is add information, if available, as to the types of transistors used in the models listed therein and link to the articles on that type. In any event its absence is not a reason to not update this article. Tom94022 (talk) 19:04, 28 May 2018 (UTC)

The time has come

With regards to adding S/360 to this list I think the time has come to do so - by my count there are 4 editors in favor, 3 not taking a postion and 1 opposed. Since we have three RS's using either "discrete," "microscopic" or "individual" to describe the transistors used in the logic of the early S/360s the opposition amounts to just WP:IDONTLIKEIT which is not a valid objection. A draft entry is at User:Tom94022/sandbox; any comments or suggestions as to form or content of the entry before I update this article. Tom94022 (talk) 19:04, 28 May 2018 (UTC)

You might want to Wikilink "hybrid circuit". Guy Harris (talk) 19:32, 28 May 2018 (UTC)

I added System/360 to the article with a simplified entry using only one of Tom's links and without the individual model numbers. I don't think we don't need to beat this to death, it's a list article. FWIW this link http://www.chipsetc.com/the-ibm-slt---solid-logic-technology.html has a video of the SLT assembly process showing the placement of an individual transistor. It looks like a modern robotic assembly line except its over half a century ago.--agr (talk) 09:53, 29 May 2018 (UTC)

Tom94022 in his recent edit, which I otherwise support, says "I'm not sure I agree with listing all S/360 in 1965 since some shipped in later years." If we list some models in 1965, then presumably we should list later models in their year of introduction, which would badly distort this list. Other computers listed had new versions introduced from time to time (e.g. GE-200 series). You might argue 360 models were more different than model changes in earlier computers and some were, but some weren't and we would end up recapitulating the entire 360 history. A list article like this is primarily a navigation tool; we don't have to get into the model issue at all. Readers who need more nuance can go to the individual articles.--agr (talk) 16:48, 29 May 2018 (UTC)

ArnoldReinhold consistancy does not seem to be a virtue in this article, note the separate listings of transistor models from the IBM 70x* series and the grouped listing of the CDC 3000 series. I corrected the GE-200 series date in the CDC style. Now that I think about it, where there is a timeline we can link to perhaps all we should do is list a series as "First model[s] of xxxxxx, see yyyyyy for later timeline models" Tom94022 (talk) 17:48, 29 May 2018 (UTC)

Listing individual models does allow us to indicate, for example, "the PDP-8 and PDP-8/S used discrete transistors; the PDP-8/I and subsequent models used monolithic integrated circuits". Guy Harris (talk) 19:09, 29 May 2018 (UTC)

Where there is an overlap, as with the PDP-8, we can simply note that. As for a timeline, here is a suggestion for a rather big project. Computers and Automation magazine included a monthly estimated census of all installed computer systems from 1962 to 1974, which covers most of the transistor era. The census sections have been collected and made available on-line at https://archive.org/details/bitsavers_computersArCensus196274_16451676 That could be turned into a Wikipedia article listing each model and its census count as of the end of each year.--agr (talk) 10:03, 30 May 2018 (UTC)

I took a hack at dealing with multiple devices and families here. The idea is to have a family on one line but to indicate the total number of models and time span on that one line, as in these two examples already posted: 1965

• PDP-8 & 8S (1965 & 1966)
• IBM System/360 family, 14 models (1965-1971). Used IBM SLT hybrid circuits.

Notes:

1. Where the number of models is three or less we just list them and parenthentically in order the year shipped
2. For families having more than 3 models we list the count of models and parenthentically the year range
3. Either a Wikilink or a link to an external source is provided to identify the models and dates.

If we agree on this format we can put a few words in the lede and then let the article clean itself up over timer. Comments? Tom94022 (talk) 16:51, 30 May 2018 (UTC)

References

Missing machines

I noticed some machines that were not in the list. In one case (CDC 7600) I'm not sure of the technology used.

• Bendix
  • G-21 - An MP version of the G-20; AFAIK only delivered to Carnegie Tech (now Carnegie Mellon University )

• CDC
  • CDC 160 series 160, 160-A, 160-G
  • 924
  • 6500 - MP with 2 6400 CPs
  • 6700 - MP with one 6600 and one 6400 CP
  • CDC 7600 - was it discrete or IC?

• DEC
  • LINC-8
  • PDP-4
  • PDP-5
  • PDP-7
  • PDP-9
  • PDP-12
• GE
  • GE-200 series GE-205 GE-215 GE-225 GE-235
  • GE-210
  • GE-400 series 415, 425, 435,[1] 455 and 465
  • 412

• Honeywell
  • rest of 200 series - 1200, 1250, 2200, 3200, 4200 and the later 2070, and the character processor of the Honeywell 8200.

• Philco computers
  • TRANSAC S-1000
  • TRANSAC S-2000 210, 211, 212 (213 never shipped)

• RCA
  • Communications Data Processor (CDP)
  • 301
  • 601 - related to CDP
  • 3301 - successor to 301

• Sylvania
  • 9400 - related to MOBIDIC

Shmuel (Seymour J.) Metz Username:Chatul (talk) 17:26, 29 May 2018 (UTC)

Here is a reference for the CDC 7600 being discrete: http://gordonbell.azurewebsites.net/craytalk/sld053.htm I still think it is overkill to list individual models, especially as we do not have dates form many examples, e.g. GE-600 series. --agr (talk) 17:44, 29 May 2018 (UTC)

The CDC 7600 is not a new model, but a new machine; the I/O architecture is very different. However, lumping the 6400, 6500, 6600 and 6700 as CDC 6000 series would make sense, other than the matter of dates. Similarly for listing the CDC 160, 160-A and 160-G as CDC 160 series. Meanwhile I forgot all about the Data General Nova and SuperNova. Shmuel (Seymour J.) Metz Username:Chatul (talk) 18:47, 29 May 2018 (UTC)

Chatul thanks for the list but if you have the dates why don't you just go ahead and add them to the list? Otherwise, to paraphrase a saying, this is just throwing stuff up on the wall to see what sticks. Tom94022 (talk) 17:48, 29 May 2018 (UTC)

Alas, I don't have most of the dates. Shmuel (Seymour J.) Metz Username:Chatul (talk) 18:47, 29 May 2018 (UTC)

The IBM M44/44X was removed. It was a special modified IBM 7044 with some form of MMU, built for research purposes (never commercialized), so I don't know whether it deserves a separate mention from the 7040/7044. Guy Harris (talk) 19:18, 29 May 2018 (UTC)

Sorry, I incorrectly removed it thinking it referred to the S/360 M44. We do have other one of a kinds so I'm going to add it back. Tom94022 (talk) 19:37, 29 May 2018 (UTC)

How to handle models

In many cases a manufacturer has several models announced and shipped on different dates but capable of running the same software. List of transistorized computers is inconsistent in how such cases are handled. I believe that the article should either list all models with the appropriate date or should list only the series with the date of the earliest model. Also, the article sometimes lumps incompatible machines together, e.g., CDC 3000 series. Some examples of basically compatible machines with different dates are

• CDC
  • CDC 3000 series
    • 3600 (June 1963), 3400, 3800 (December 1965)
    • CDC 3200 (May 1964), CDC 3100 (February 1965), CDC 3300 (December 1965)
• GE
  • 615, 625, 635 (645 had paging)
• Honeywell
  • Honeywell 200 series - 200, 1200, 1250, 2200, 3200, 4200 and the later 2070
• IBM
  • 1401 1440 1460
  • 1410, 7010
  • 7040, 7044
  • 7070, 7072, 7074
  • 7090, 7094, 7094 II
  • System/360 - multiple models

-Shmuel (Seymour J.) Metz Username:Chatul (talk) 19:23, 29 May 2018 (UTC)

This list is not particularly helpful. It would help if you took one or more of these families and prototyped the possible ways as you think could be displayed in the article. Tom94022 (talk) 21:57, 29 May 2018 (UTC)

Clary DE-60 refs

More Clary DE-60 references for people creating article about it

"Old Calculator Web Museum Calculator Timeline". www.oldcalculatormuseum.com. 1957 Jan, 1960 Feb, 1961 Mar.

• "Old Calculator Web Museum - News Archive - Clary DAC-2500 Digital Arithmetic Center Announcement". www.oldcalculatormuseum.com. Cited article doesn't include last paragraph (Size: [...]), full article below.{{cite web}}: CS1 maint: others (link)
  • "News of Computers and Data Processors: ACROSS THE EDITOR'S DESK". Computers and Automation. 10 (3(B)). 196103.pdf: 9B (33). Mar 1961.{{cite journal}}: CS1 maint: others (link)

American Data Processing, Inc. (1962). "100-101 CLARY DE-60 Clary Corporation". Data processing equipment encyclopedia. Supplement. 1 (3). American Data Processing, inc.: 6–7. {{cite journal}}: |first1= has generic name (help)

"1959 PICTORIAL REPORT ON THE COMPUTER FIELD". Computers and Automation. 8 (12). 195912.pdf - Pictures of computer: 9. Dec 1959.{{cite journal}}: CS1 maint: others (link)

• "1961 PICTORIAL REPORT ON THE COMPUTER FIELD: 1. Digital Computer". Computers and Automation. 10 (12). 196112.pdf: 25. Dec 1961.{{cite journal}}: CS1 maint: others (link)

"ACROSS THE EDITOR'S DESK: RACE TRACK USES PORTABLE COMPUTER". Computers and Automation. XI (2). 196202.pdf: 30. Feb 1962.{{cite journal}}: CS1 maint: others (link)

--89.25.210.104 (talk) 22:12, 24 August 2018 (UTC)

If you think they're important, you should add them yourself; Wikipedia is, to quote the front page, "the free encyclopedia that anyone can edit", which means that the right way to change a page is, ultimately, to edit the page, not to edit the talk page and say "would somebody please do this to the page?"

If you're not sure what should be done to the page, the appropriate thing to do would be to ask, on the page, "should the page be changed in this fashion?" or "what should be done to the page to address this issue?" and, once a consensus is reached, making the change agreed on by consensus. Guy Harris (talk) 05:32, 25 August 2018 (UTC)

They were added, but were removed ("redundant poorly formed/formated refs") - so I put them here as a backup (otherwise they would be lost in the page revision history) in hope that they will be of some use for someone else. --89.25.210.104 (talk) 21:57, 26 August 2018 (UTC)

Would it be better to move those references to the talk page of a Clary DE-60 (and make redlink on the list page)? --89.25.210.104 (talk) 22:35, 26 August 2018 (UTC)

Why don't you start an article on the Clary DE-60? You can start with a stub including all the references. You could then change it in this article from a ref to a link.

"redundant poorly formed/formated refs" was a summary for as far as references go on this page one should be enough to establish the facts (date and transistorized); a large number of redundant references clutters the article. Multiple references usually are usually formatted into separate tags rather than multiple bulleted links under a single tag. Tom94022 (talk) 17:05, 27 August 2018 (UTC)

I suspected someone would say something like that :) Maybe I do just that.

Multiple references (over 3): usually, but bulleted citation bundling also is an option. --89.25.210.104 (talk) 23:57, 27 August 2018 (UTC)

More potential additions

There is a list of 2^nd generation computers at second generation that should be added to this list. The dates and technology have been confirmed from a variety of RSs, many of which are linked or referenced by the manufacturer/model listed in the table. At first glance none of them are directly listed in this article (a few may be indirectly listed such as CDC 3800). It's a project I don't have time or inclination to take on, so if someone following this article does, please do. Tom94022 (talk) 17:23, 3 April 2019 (UTC)

subsections

Should the article have subsections for each year, or would it be better to replace the years with the semicolon header? I favor replacing the subsections with the semicolon header. Bubba73 ^{You talkin' to me?} 01:47, 20 October 2019 (UTC)

I'm not sure there is any substantial benefit of a change so unless there is one there is no reason to change. Personally I am content with subsection for each year, see: WP:IDONTLIKEIT. Tom94022 (talk) 06:05, 20 October 2019 (UTC)

Dates and sort order

I wanted to add some additional machines and I realized that there is an inconsistency in the dates: some, e.g., IBM 7090, have the announcement year but the lede says that the list is by delivery year.

According to IBM 7090 its installation was Dec 1959 so its placement in this article is correct. Tom94022 (talk) 16:49, 23 August 2020 (UTC)

There seems to be a conflict between two sources, oneclaiming December 1959 and another claiming June 1960. I have a bunch of machines that I wanted to add, and I was counting on the Adams quarterly as a source for the delivery dates. Should I trust them for the other machines, or hold off? Shmuel (Seymour J.) Metz Username:Chatul (talk) 05:26, 25 August 2020 (UTC)

The are both reliable sources and the disagree. IMO IBM is more reliable about IBM's dates than Adams so I would leave the 7090 where it is. It's up to you as to whether u want to footnote the Adams date or just ignore the difference. And just because there is a difference doesn't mean Adams is not a RS so I would use it for other models and if there are differences apply similar judgements. These are differences not disputes. Tom94022 (talk) 06:15, 25 August 2020 (UTC)

I found an article in Datamation that agrees with the IBM source, so I'll leave the 7090 entry as is. Or should I add the month there and in the new entries? Shmuel (Seymour J.) Metz Username:Chatul (talk) 11:48, 25 August 2020 (UTC)

It's your choice - I would Tom94022 (talk) 21:28, 25 August 2020 (UTC)

Also, is there a preferred sort order within year? Shmuel (Seymour J.) Metz Username:Chatul (talk) 13:48, 23 August 2020 (UTC)

Interesting question - in many cases the month may not be known. There seems to be no order. IMO it probably should be alphabetical in a year but that is a major task and should require consensus before started so I suggest u just add the new material at the end of a year. If you want to alphabetize a section you add to that's OK also. Tom94022 (talk) 16:49, 23 August 2020 (UTC)

References

CDC 924A

What connection does it have with the CDC 924? Guy Harris (talk) 07:02, 30 August 2020 (UTC)

The 924A was an enhanced version of the 924. The 1604A, 1604B and 1604C were enhanced versions of the 1604. The 160A and 160G were enhanced versions of the 160. The 160G had an extra bit but ignored it in compatibility mode.

BTW, do you have dates for the 3150 and 3400? Shmuel (Seymour J.) Metz Username:Chatul (talk) 15:12, 30 August 2020 (UTC)

The Adams and Associates Computer Characteristics Quarterly for first and second quarters 1967 says, on page 14, that the 3400 was first delivered in November 1964. It doesn't list the 3150. However, the fourth quarter 1967 - first quarter 1968 version lists the 3150, on page 11, as being first delivered in 1967, no month specified. Guy Harris (talk) 21:46, 4 September 2020 (UTC)

Recent reverted edit

@Tom94022:I just reverted an edit because it created too many redlinks and inappropriate links. E.g., in RCA 601 there is no RCA 601 article, in UNIVAC II, III the first link should be to UNIVAC, not to UNIVAC 1004 and the second link is redlined. Either [[UNIVAC 1004]] I, II, III or [[UNIVAC]] [[UNIVAC 1004|1004]] I, II, III would work.

With regard to delivery dates, some sources are:

• First and Second Quarters 1967, COMPUTER CHARACTERISTICS QUARTERLY, vol. Volume 7, Numbers 1 and 2, adams associates {{citation}}: |volume= has extra text (help)
• "News Briefs in Datamation 1st 601 is on the air at NJ Bell" (PDF), Datamation, vol. 9, Number 6, GARDNER F. LANDON, p. 52

Shmuel (Seymour J.) Metz Username:Chatul (talk) 08:53, 3 September 2020 (UTC)

Chatul With all due respect if there are errors please fix them rather than revert the entire change which I believe is for the most part accurate.

• Regarding redlinks, they are appropriate for information for which there is no reliable source. Actually if there is no reliable source then the item should not be listed at all but a redlink is one way to establish the need for an RS. I suppose I could have used a dubious tag.
• Regarding the two cites above, if they are RS's for inclusion of an item in the list then please link them in the list. They don't do much good here.

I will look into the two items u mention above, but I am going to revert back to what I posted before hand. Please do not revert again but take the time to improve the article. Tom94022 (talk) 16:33, 3 September 2020 (UTC)

There are lots of RSs for an RCA601 but no article and no RS for a date so I left it as a redlink. Tom94022 (talk) 16:45, 3 September 2020 (UTC)

As it happens there are articles for both the U II and the U III with a date for the U III so I deleted the offending and incorrect line. There is no date for the U II in the article so if you are interested in improving Wikipedia why don't you go find a date, add it to the U II article and then link to it from this article. Tom94022 (talk) 16:54, 3 September 2020 (UTC)

First, I started editing your change and realized that there was so much wrong that it was easier to revert and add back the few changes that were legitimate. Second, I provides sources for all of the dates on this talk page. In particular, why didn't you look at the second ("News Briefs in Datamation 1st 601 is on the air at NJ Bell"PDF) citation before pretending that the 601 didn't exist? You didn't even need to follow the link to notice that there was a source for it. Shmuel (Seymour J.) Metz Username:Chatul (talk) 07:03, 4 September 2020 (UTC)

You didn't add back anything and as near as I can tell there were few if any errors in my edit - the two you asserted were not. You did not like the way I identified those edits that did not have a RS's for dates but that is a matter of style not correctness. You do know that RS's are required for inclusion in a list don't you? In the end you added in a number of RS's which is what you should have done in the first place. Regarding the 601, may I again suggest that when you have changes to make to an article, you make them in the article rather than giving assignments to other editors. But thanks for all the recent work, the article is now much improved. Tom94022 (talk) 16:59, 4 September 2020 (UTC)

You added a bunch of redlinks that I had to clean up and you piped product links to the name of the manufacturer. Both created extra work for me. I would have completed the updates in half the time had you not reverted my reversion.

You do know that most of the entries on the list don't show a RS, don't you? Are you planning to delete the rest of them instead of flagging them?

Please don't misrepresent what I didn't like. I might have preferred {{cn}} to <!--no RS for date--> but I never mentioned that. What I objected to was breaking links to manufacturers and adding redlinks.

May I suggest that what you should have done in the first place is to add {{cn}} tags, and that what you should have done when the edit summary saif see talk page was to discuss it on the talk page instead of edit warring? Shmuel (Seymour J.) Metz Username:Chatul (talk) 21:53, 4 September 2020 (UTC)

CDC 1604B date

Chatul in the 1963 section you have depicted the 1604B as "[[CDC 1604]]B"; however, I cannot find neither the model nor the date at CDC 1604 nor after a reasonable search can I find an RS for 1963. I have found several RS's as to its existence but nothing as to the appropriate date section in which it should it be listed. I believe the reader has a right to know that there may be an issue with this entry. It seems to me I can take one of three actions, in all cases stating in the edit note that the entry lacks an RS as to date:

1. Delete the entry.
2. Mark the entry with a dubious tag, stating the reason as it lacks an RS as to a date
3. Change the entry into a redlink.

An editor who finds a RS can update the article as appropriate. Personally I like #3 as an interim solution since it acknowledges the good faith of the editor that entered it and is more compact than #2. Of course, for both #2 and #3 after a period of time without an RS the entry should be removed.
You seem to have strong feelings on this issue; how would u have me mark up the article? BTW, if you have an RS please go ahead and edit the article but I am still interested in how you think such situations should be marked. Tom94022 (talk) 22:14, 4 September 2020 (UTC)

Well, this is amusing. For what it's worth, this oral history of Edward S. Davidson (1939- ) says:

00:12:04 ED: Well now I had to write a program. And although I had taken programming, I wasn't particularly good at it. We had a computer in the lab which was a Control Data CDC 1604B. It was a long time before I knew that Control Data actually never manufactured the 1604B. They did a 1604 and a 1604A, and we had a engineer on staff who managed the computer installation. If I recall his name was Ernie Neff, N-E-F-F. And he took a field service training course at Control Data, he was licensed to be a field service engineer but he was employed by our lab, Coordinated Science Lab. He was an engineer's engineer, a putterer, he was our own kind of guy. So when he was bored, which he often was, because he didn't have an awful lot to do, he would look at the machine and say "Gee, I could optimize that circuit a little bit." He'd start rewiring circuits inside the machine, which you could do with wire wrap in discrete packages you could slip in and out. And he minimized, minimized and rewired and packed it in and pretty soon he got a couple of empty slots. So he ordered some new spare cards and he'd add a feature that had never been part of the original machine.

00:13:37 PE: So you had a custom computer?

00:13:38 ED: Well not only a custom computer, but a custom undocumented computer. Which was very interesting. ...

That sounds as if it was at the University of Illinois. However, this issue of Computers and Automation lists, on page 131, the University of Wisconsin as having a "CDC 1604B/160", and this student newspaper from Northwest Nazarene College in Nampa, Idaho says, on page 5, that "Professor LaVerne Rickard, Director of NNC's Computer Cen­ter, has announced that a govern­ment agency with a higher priority has preempted our claim on a CDC 1604b computer." I dunno, maybe they sent the changes back to CDC who picked them up and started making a 1604B? Guy Harris (talk) 22:45, 4 September 2020 (UTC)

Oral histories can be unreliable and there is ample evidence of the existence of the 1604B but no good reference yet for a date - so what do we do? BTW, this source suggests it might be 1964 or earlier. One thing we should do is add something about the 1604B to the 1604 article but so far I don't think we have an RS for listing it in this article. So what do we do? I suppose we could list it in 1964 with an "or earlier" caveat using the found source.

Chatul FWIW I intend to mark other elements without an RS for a date with a redlink. If you have a better idea, i'd like to hear it. Tom94022 (talk) 06:18, 5 September 2020 (UTC)

I've added {{cn|reason=Need RS for date.}} on 1604B; that flags it without creating extraneous work for other editors. I thought that http://ed-thelen.org/comp-hist/AdamsReport1967Q4-1968Q1.pdf might have the missing dates, but no such luck.I did find, and add, a source for the 1604 date. There are manuals for a 1604C on bitsavers, but I don't know the dates.

I initially added sources for dates as [URL], but user:Guy Harris edited them to footnotes using {{cite}}, and I think that looks cleaner. Shmuel (Seymour J.) Metz Username:Chatul (talk) 04:06, 6 September 2020 (UTC)

Personally I think [URL Product Name] is the cleanest for an EL but we should use a consistent method for citations within an article and it looks like reference with citation is what we have. {{cn|reason=Need RS for date.}} works for date needed but right now there is no existence proof for the 1604B at CDC 1604 so we still have a problem with this entry. Tom94022 (talk) 07:06, 6 September 2020 (UTC)

UNIVAC Solid State 80 and 90?

My understanding is that the [UNIVAC SS 80 and SS 90 mostly used magnetic logic rather than transistors, although it did use magnetic amplifier, transistors and vacuum tubes. I would classify it as solid state but not transistorized. -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 16:18, 31 January 2024 (UTC)