Talk:CDC Cyber

Too technical?

{{Technical|date=March 2008}} (Tag disabled on purpose)

Regarding the above: "You should put an explanation on the talk page with comments on why you believe it is too technical, or suggestions for improvement. Templates added without explanation are likely to be either ignored or removed." Funkymonkey (talk) 15:44, 3 November 2010 (UTC)

Although the March 2008 version may have been too hard to understand, since then it has been completely rewritten. The current version is fine, so I am removing the technical tag with NOWIKI. Guy Macon 18:19, 8 February 2011 (UTC)

Untitled

Wasn't there a Cyber 70 series as well? I seem to remember a friend showing me listings with that name.

• Yes. IIRC, the 170 was line the 70 series except that it had something extra - some extended memory or dual processor - something like that. (I can't quite remember.)

--Bubba73 02:44, 3 Jun 2005 (UTC)

The 70 series was a slightly upgraded 6000 series, still no parity checks anywhere in the mainframe. The 170 series introduced parity; the 170/700 series introduced SECDED (singled error detection, double error correction) plus the SCR (status control register) for hardware monitoring.206.10.110.217 04:56, 26 November 2005 (UTC)Mike Vollmer

As noted above, the Cyber 70 series was basically the same as the earlier machines. The 6400 became a Cyber-73, the 6600 became a Cyber-74, and the 7600 became a Cyber-76. The 6400 got a few new instructions for character processing, called a CMU (Compare/Move Unit.)

The Cyber-170 series was a major technological change: Chips were used instead of individual transistors. And solid-state memory was used instead of core memory. The Cyber-170/700 series was a technology refresh of the 170-series.

The Cyber-180/800 series was also a major change. Most of the 800-series machines could run in a "dual-state" 60-bit and 64-bit mode. The 180 machines, when run in 64-bit mode, used a new operating system called NOS/VE.

The Cyber-200 series machines were 64-bit machines based on the STAR-100. They were 64-bit vector processing machines which had little in common with the "traditional" machines.

Wws 6/4/2006

Cyber 2000

There was also a CYBER 2000. I believe it was a 64-bit-only machine that ran NOS/VE and not NOS. As of 2006, FIDES Information Services (which I believe is in Switzerland) was still running one. I will see whether I can get CDC experts Fredy Ferrari or Markus Mueller to weigh in on this. Riordanmr 21:58, 28 January 2007 (UTC)

Operating system

"The systems typically ran CDC's NOS (Network Operating System), 1.4 or 2.0. Several other operating systems were available for the machine, but the only one to see any real use on the 170's was NOS/BE" I seem to remember KRONOS being used, but that may have been on the earlier 8000 line. Anyone know? --Bubba73 02:42, 3 Jun 2005 (UTC)

I changed the reference to NOS 1.4 or 2.0 since there were other releases of both OSes. As to the question about Kronos — Kronos was a predecessor to NOS, but darned if I can remember the name of predecessor to NOS/BE. -- JonRoma 06:38, 22 August 2005 (UTC)

NOS/BE's predecessor was Scope. Updating accordingly -- JonRoma 08:07, 25 August 2005 (UTC)

SCOPE was the original OS on CDC's 6000-series mainframes.

KRONOS started as a benchmarking system written by CDC's software guru Greg Manfield and later marketed; in time it was renamed NOS. I used to use a KRONOS Analysis manual when I worked with a NOS (1.3?) system in the late 70's; it was sufficiently current.

NOS/BE was SCOPE with its user interfaces revamped to look more like NOS.

In the end, SCOPE was retained only for the 7600, which did not support the PPS required for NOS & NOS/BE.

From my NOS Analysis instructor in the summer of 1982: SCOPE = Sunnyvale's Collection of Programming Errors, NOS = No Operating System, NOS/BE = No Operating System/Bad Example. 206.10.110.217 05:08, 26 November 2005 (UTC)Mike Vollmer

Originally, the CDC 6000 series was intended to run an operating system called SIPROS (SImultaneous PROceSsing). SIPROS was late, so systems were actually delivered with a version of the Chippewa Operating System, COS, which Mr. Cray and his team used for system checkout and diagnostics. SCOPE evolved from COS and was widely used. Some time later, Mansfield and team once again took COS as a base and created MACE - which was both faster and more configurable than SCOPE. MACE evolved into KRONOS - which supported time-sharing. At this point, CDC had two operating systems - one largely tied to batch processing, the other for time-sharing - which were somewhat, but not entirely, compatible. When the Cyber-170 systems were released, circa 1975, KRONOS was renamed NOS and SCOPE was renamed NOS/BE. CDC then placed most of their 60-bit development emphasis into NOS.

Note that the 7600 and Cyber-76 ran 7600-specific versions of SCOPE.

Wws 06/04/2006

170 address registers

Seven of the A registers were tied to their corresponding X register. Storing an address into register A1, for example, caused the contents of the central memory word referenced by that address to be fetched into register X1. Likewise, storing an address into register A7 caused the contents of register X7 to be stored to the central memory location referenced by A7.

Storing into A1 fetches into X1, but storing into A7 stores from X7? Some of the seven fetch and some store? If someone would fix or clarify… -- RandallJones 06:22, 25 August 2005 (UTC)

Yes. The CDC Compass Version 3 Reference Manual publication no. 60492600 revision G, dated July 7, 1980 states:

8.3.2 A Registers

Eight 18-bit X registers in the computation section of the CPU, designated as A0, A1, ..., A7, are essentially SCM (CM) operand registers. With the exception of A0 and X0, A registers are associated one-for-one with the X registers. Placing a quantity into an address register A1 - A5 causes an immediate SCM (CM) read reference to that relative address and sends the SCM (CM) word to the corresponding operand register X1 - X5. Similarly placing a value into address register A6 or A7 causes the word in the corresponding X6 or X7 operand register to be written into that relative address of SCM (CM).

The A0 and X0 registers operate independently of each other and have no connection with SCM (CM). A0 is used as the relative SCM (CM) starting address in a block copy operation and for scratch pad or intermediate results.

and also

8.4.45 Set A Register Instructions

These instructions are intended for fetching operands from storage for computation and for delivering results back into storage. The instructions have two destination registers: the Ai register, which receives the address formed from the operands, and either the Xi register or a CM (SCM) storage location.

...

The type of storage reference is a function of the i deignator value, as follows:

i = 0; no storage reference

i = 1, 2, 3, 4, or 5; contents of CM (SCM) relative address (Ai) to register Xi

i = 6 or 7; contents of register Xi stored at CM (SCM) relative address (Ai)

Editor's note: Note that "SCM" refers to "semiconductor central memory" (the type of storage in this generation of Cyber systems) while "CM" is the generic term "central memory" used for the entire architecture from the CDC 6000 forward.

No, SCM refers to Small Central Memory in 7600/7700s and Cyber 76s. Compare LCM, Large Central Memory. These correspond roughly to CM and ECS (External/Extended Core Storage) in that LCM was not external, but the instructions for accessing it were the same as ECS instructions on other machines, plus a few more (also present in 180s, where it's CM and UEM. ECS/ESM for 180s was designed as a special option). The Cyber 176, being a high-speed collision between a 170 and a 76, has CM and LCME, PPs and PPUs... See manual 60420000H for details. Jlwwp (talk) 20:53, 12 February 2022 (UTC)

I omitted this specifics in the article, since it duplicated information available elsewhere and since there were numerous other hardware details that I could have added (and probably will add at a later date) — along with some pictures.

I can clarify.

These machines were most interesting to work on as a youngster of age 19 back in the "good old days". -- JonRoma 08:02, 25 August 2005 (UTC)

A reference for COMPASS?

Somewhere in this article there should be a reference for COMPASS. -- FedKad 06:20, 26 August 2005 (UTC)

UTexas

Don't know if it belongs in the article, but well into the 1990s (maybe into the second half of the 1990s), the University of Texas at Austin ran computer account billing and accounting functions on a "CDC Dual Cyber 170/750" running the custom "UT-2D" operating system. Churchh 17:51, 24 April 2006 (UTC)

I doubt it belongs in the article, but likewise I will never forget my access to the CDC Cyber when I went to college at Georgia Tech in the late 1980's, my first taste of using a "real computer". I learned Unix on it. Daniel Freeman (talk) 12:26, 23 May 2008 (UTC)

Byte?

I don't recall encountering the word "byte" until after Cybers and CDC 3300s (the first were PDP-11s and Motorola 6800s). Does someone actually recall CDC lingo using "byte"? I do remember that every single CDC manual had a Display Code appendix—which were well worn. EncMstr 02:34, 19 May 2006 (UTC)

I've seen CDC docs where they use "byte" to refer to 12-bit quantities (each containing two 6-bit characters.) This makes sense when one considers that the PPs and I/O architecture of the machine had a 12-bit orientation. Wws 6/4/2006

Yes, at Michigan State Univ (a CDC shop), we used the word "byte" to refer to the 12-bit PP words. I hated the use of the word that way, but that was how it was used by our systems programmers, and apparently by CDC programmers at other shops. I was exposed to the use of "byte" as 12 bits probably during the local Systems Programmer class in 1980, but it had apparently been in use for many years prior to that. Riordanmr 21:47, 28 January 2007 (UTC)

The definition of byte and word was computer dependent until the 80's. I went to Georgia Tech back in the 70's and 80's. As a freshman, I programmed in Fortran on the Cyber and as I got older, I moved into the lab where I programmed microprocessors. Each machine had different sized bytes and words. In my memory, the definition of byte didn't stabilize at 8 bits until two things happened, the Z80 and 8080 based personal computers became popular. Then, when IBM came out with one, it was essentially carved in stone. The other thing was ASCII. At first, at least for me, computers were for math and control. But, increasingly, word processing on general purpose machines became popular. ASCII characters fit very conveniently into 8 bit bytes. If the registers were bigger than 8 bits, it was easy to make their size multiples of 8. So, word was useful for describing register size and address space as well as word processing. It meant almost the same thing, a meaningful collection of 8 bit bytes/characters. Siriuskase (talk) 16:00, 15 October 2010 (UTC)

CYBER vs Cyber

Yesterday I discovered that user "Yworo" changed all instances of "CYBER" to "Cyber". Control Data Corporation consistently referred to the mainframe as "CYBER". I have attempted to fix this, but Wikipedia "administrators" repeatedly undid my fixes.

The "Desktop CYBER" page was also edited by "Yworo" and later the link to the Subversion repository of the emulator was also removed by "Orangemike".

Does anybody else care about the correct spelling of "CYBER" and the integrity of links?

If yes, please check out the history page of "CDC CYBER" and "Desktop CYBER" and let us know what you think.

Cdccyber (talk) 11:38, 5 August 2011 (UTC)

You are incorrect. CYBER with all caps is a styled trademark. According to WP:ALLCAPS, trademarks styled in all caps should be reduced to normal capitalization. So should article titles, which is why this article is titled "CDC Cyber" and not "CDC CYBER". This is a style issue and we do not keep all caps unless they are an acronym. Sorry, please familiarize yourself more with our manual of style. Yworo (talk) 14:50, 5 August 2011 (UTC)

User Cdccyber claimed that Control Data Corporation consistently used the name "CYBER" in all caps and he is correct despite any Wikipedia stylistic conventions to the contrary. Please familiarize yourself with the subject before pronouncing the claim incorrect. — Preceding unsigned comment added by 74.196.171.227 (talk) 15:29, 20 February 2014 (UTC)

Edit to Cyberplus dimensions list

To explain myself, as the full reasoning wouldn't fit into the space available.

First up, even on seeing the dimensions as they were, I was somewhat skeptical. About 3.5 to 4.5 metres wide, by 1.6 deep, by 4.9 high? That's a huge box; for those of you who can't think in metric, that's about 11'5" ~ 15'3" W, 5'3" D, and 16'1" H. Or in other words, three subcompact cars stacked on top of each other (and probably microvan styled ones, at that, as they wouldn't be resting on their wheels...).

Computer cabinets, even supercomputer mainframes, just don't come in that kind of size as far as I'm aware. It'd require a specialised building to site them in (plenty enough places barely have 8-foot ceilings as it is, before you start adding raised floors for wiring and cooling systems, so you'd have to knock three storeys together, not just two - as well as making the doorways between the delivery dock and the actual operating area rather outsized), or at least something more akin to a warehouse than a regular computer room or research centre, and would make delivery, commissioning and service/maintenance very difficult as the things would be way beyond normal "human scale"... you'd need ladders to get at some parts, and there's a risk that techs may get lost in the workings depending how it was laid out. The delivery might also be a touch conspicuous as a result, unless you're going to lay them flat in a container or maybe take them two at a time (still laid sideways) in a boxvan. Not what you want for black ops.

Sure, it'd be a hell of an impressive monolith to scare new recruits with when giving them the tour, but other than that, such behemoths belong to the world of sci-fi. Long or wide? Sure. Really tall? Less likely. Really tall in combination with being kinda wide, yet only moderately (maybe two filing cabinets) deep? Not buying it, sorry.

It's inherently a modular system, and developed partly as a result of the company's drive towards miniaturisation and VLSI, so why would it end up larger than any single piece of pretty much any computer system that came before it, including those from the relay-and-valve days?

Given that most systems followed the IBM model of being very human in their module size - with the largest individual parts still just about fitting through a standard doorway, and being able to turn a corner in a regular corridor - surely this excessive size would also be considered noteworthy amongst anyone commenting upon the system elsewhere? But no, no mention of how large it was in basically every other source amongst the limited selection that can be found online (it really was quite the blackops thing, there's not much info out there... but there is SOME). In fact, the only one which made any reference to it at all was that Usenet post I cited, from way back around Christmas 1991 - at a time when the system itself would still have been in use and very relevant, possibly even still being installed, because the power and storage it offered, at least in cluster format, would still have been quite impressive by the standards of the day.

Interestingly enough, the contributor to that thread who makes mention of the system, and states it as being the same size as a Cyber 180 (enough pictures of which exist to show that it's a perfectly normal sized rig, at least in mainframe terms; no taller than the average man, no deeper than you can comfortably reach with your arm, and only the more low-slung components being wider than your fingertip-to-fingertip arm span, and even those didn't look more than 3 metres long)... has the same first name as the Wiki contributor who added the questionable information some 8 years ago, in early 2006. Now, it could be just a coincidence, seeing as "Gunther" isn't that rare a name, at least in Germany, but it does pique the interest when seen in combination with such a niche piece of computing equipment.

As slanderous as it might be to suggest it, might it be that Gunther's memory faded somewhat in the fourteen years and one month (almost to the day!) between the relatively fresh hands-on report they made in the early 90s, and their contribution of a bunch of uncited figures (measured how? they're not even mentioned in the limited CDC documentation I scared up) to this page...? :-)

I'm trying to figure out, even, if there might be some units confusion, e.g. perhaps they're actually 3'5" x 1'6" x 4'9"? And that somehow morphed to 3.5m/etc and then to "350cm"? The actual figures given are too precise, though... they're not actually rounded off to the nearest 10cm or whatever. Confusing. Where did the numbers come from? I also suspected that Height and Depth may have been transposed, but despite it being a more sensible height (equal to the shoulder height of an average man), that's still a quite large monolithic block to have to reach into or crawl around inside of; again, like three small cars parked tightly together - what happens if you need to replace the handbrake handle release button in the one in the middle of the sandwich? And just how, pray tell, do you deliver the blasted thing? Unless it's on castors, delivered to a no-step loading dock, and has a clear, wide run all the way to the actual data centre (going through, one presumes, multiple security posts along the way).

The only last option is that it refers to the overall size of the cross-shaped configuration that they supposedly formed up into, and non-FPU models only formed a "T" instead. But those still wouldn't have been transported as-built for obvious reasons (instead, the 3 or 4 individual boxes surely would have been disconnected and packed together)... and it ends up counting all the empty space between the arms, and within the inner square (if there was one?) as part of the machine, and doesn't give us any clues as to how thick the arms themselves were.

The other thing to consider is the weight of the thing. Something that measures about 27 to 35 cubic metres but weighs ONLY one tonne is surprisingly low density for computer equipment. If you sealed it up watertight, it would make a very effective floatation aid. I'd be surprised, in fact, if the frame itself weighed under 1000kg before you put any electronics inside it, and much more so if it was under 250, leaving the other 75% of the total load free for actual computer parts. Parts that, going by how CDC built the other models mentioned in the article, were themselves rather higher density than most, using heavyweight 1980s components, heavy solid metal heatsinks built right into the card modules, lots of cabling (thanks to the crossbar setup) and with heavyweight period power and cooling equipment. A regular modern server rack or an old mainframe cabinet weighing a quarter tonne each, believable. This being a great many times larger and only weighing a thousand? Naw, man.

Consider a typical laptop, the lighter types of which may weigh one kilogramme, and measure about 30 x 20 x 4cm, or roughly 0.0025 (1/400th) of a cubic metre. That's 2.5x HEAVIER* than water, nevermind being thirty times LIGHTER*... it just doesn't wash. Or for a maybe more closely applicable example, a miditower computer case... maybe 20cm wide, 40cm deep, 60cm high, weighing 5kg with only a minimal amount of kit inside, or 10kg when fully equipped. That's still 3 to 6 tonnes when scaled up to the stated dimensions... and I'm erring on the low density side for it, too (probably actually heavier and smaller). I've a colleague who's shipping a load of old desktops to Africa as part of a charity mission, and he's having to put them all on a pallet and contract the job to a specialist haulage company because although they'd all fit into a normal box van just fine, the thing would easily end up exceeding the weight he's licensed for... even if driving the van to mid-Africa was that practical an idea.

(* yeah ok... "denser" and "sparser").

(I have considered that something with all that heavyweight, for-the-time high tech circuitry and associated power/cooling systems, in something the same size as a larger Cyber 180 cabinet may well weigh a full tonne... hence I've left that part in...)

All the same, I'd love for someone to come and conclusively prove one of us wrong, even if it turns out to be me on the losing side. Who knows, maybe it WAS that big (and low density)? Blackops guys do some weird stuff. Just look at the F117A, which technically speaking shouldn't be able to fly at all, and falls from the sky if its avionics fail. If anyone has an actual commissioning guide for the AFP that includes its measurements, or has now / has had access to one and a measuring tape and found out how big it REALLY is, and can contribute in some kind of even halfway verifiable way, that'd be awesome. 193.63.174.211 (talk) 16:05, 6 February 2014 (UTC)

Cyber 70/ models

As mentioned above, several years ago, there was a Cyber 70/73 (based on the CDC 6400), Cyber 70/74 (based on the 6600), Cyber 70/76 (based on the 7600). That needs to be included in the article with a reference. I think there was also a Cyber 70/75, which I suspect was based on the 6500, but I don't know. (It might have been a dual 70/74.)Bubba73 ^{You talkin' to me?} 01:35, 7 January 2015 (UTC)

Etymology?

With what connotations did CDC use the name "Cyber"? Normally I wouldn't have thought of "cyber" being commonly associated with computers until at least the 80s or even 90s. Dolda2000 (talk) 03:33, 29 February 2016 (UTC)

It is not mentioned that they had 3-address CPUs and a suggestion on how to improve the article

There are 0-address or stack CPUs, 1-address with one accumulator, 2-Address with two and 3-address, with three accumulators. That mean that two operands can be placed in two accumulators or registers, and the result is placed on the 3rd.

Nothing is said about their representation for integers and floating point numbers. I vaguely remember that they had +0 and -0. But I never used any Cyber. I had that reference from some programming text, or maybe commented by friends that programmed in some Cyber mainframe, reserved for the tough guys needing intensive computation.

Nothing of that is mentioned, if it is I didn't see. The article should be rewritten with a better structure which describe the general architecture, before going to many details, because one loses the point.

Take into account, that the vast majority of readers, had never used one of such expensive mainframes. The actual reference is maybe one Intel or ARM PC. Under such context, what does channel means? Got it? the article goes to many details, before giving the whole picture. In that way, it is confusing how the CYBERs evolved and why the company disappeared. — Preceding unsigned comment added by 2806:107E:C:F09:218:DEFF:FE2B:1215 (talk) 07:15, 28 September 2018 (UTC)

The CDC machines in question are ones' complement. It does seem that should be in the article. Gah4 (talk) 11:40, 28 September 2018 (UTC)

Splitting proposal

CDC Cyber#Cyber 200 series should be split into a separate article called CDC Cyber 200. The present situation confusingly and incorrectly treats a myriad of different computers as a single topic based on CDC's convention of naming all its computers "Cyber". The Cyber 200 series are technically and commercially distinct from all other computers with the Cyber name. The Cyber 200 is distinct from the three other groups of computers described in this article: the mainframe/scalar scientific computers that descended from the earlier scalar supercomputers starting with the CDC 6600, the attached processors (accelerators), and minicomputers. The Cyber 200 are vector supercomputers, not "mainframe-class supercomputers" whatever this supposed to mean. They were marketed to scientific users, not businesses for commercial applications. There is more than enough in-depth coverage of the Cyber 200, whose models are most commonly referred to as the Cyber 203 and Cyber 205, to justify a dedicated article: I have three at hand:

• Hwang and Briggs' Computer Architecture and Parallel Processing (1984), pp. 280–293.
• Hockney and Jesshope's Parallel Computers (2nd ed., 1988), pp. 155–185.
• Ibbett and Topham's Architecture of High Performance Computers, Vol. I (1989), pp. 160–177.

This is a total of 60 pages of material, and is by no means all that has been written on the Cyber 200. HTW217 (talk) 14:42, 18 November 2021 (UTC)

I agree. Bubba73 ^{You talkin' to me?} 16:40, 18 November 2021 (UTC)

170 character sets???

"Due to the strong dependency of developed applications on the particular installation's character set, many installations chose to run the older operating systems rather than convert their applications. Other installations would patch newer versions of the operating system to use the older character set to maintain application compatibility"

I am at a loss here. I am not an expert, but have experience, and am doing a documentation project for dtcyber, and what character sets are being referred to here that required some form of "porting"?

The last version of NOS, 2.8.7 871/871, supports 63/64 character 6-bit display code sets, ASCII 128 char set, ASCII 63/64 char sets, ASCII 95 char set, and could convert 6 and 8 bit versions of EBCDIC. All of that in 6-bit, 6/12-bit, and 7/12-bit. (Appendix A, CDC 60459680R NOS Version 2 Reference Set, Volume 3, System Commands; Dec. 1997)

Whereas, the first KRONOS manual shows display code, Hollerith, and external/internal bcd. All of these codes are supported all the way through NOS2. (Appendix F, CDC 59150600A KRONOS Batch User's Reference Manual; Feb 27, 1970).

The 1968 SCOPE 3.1 Manual shows the same as the KRONOS manual, omitting internal BCD. (Appendix A, 60189400B)

The UMich SCOPE/HUSTLER manual from 1981 also gives 026 and 029, both supported all the way to NOS2...

Unless someone can show widespread porting of depreciated charsets from CDC to CDC, I suggest that the sentence be modified to reflect importation of never-supported charsets instead of reviving the depreciated within CDC. Hell, NOS2 still supported reading Univac 1103 tapes, fer reel (pun totally intended)! I could go on doing this with the CDC manuals, but why? CDC continued support of its charsets as far as 60-bit went. 2603:300B:133D:D000:45CF:A192:D820:C93A (talk) 16:23, 15 October 2023 (UTC)