Talk:IBM System/360 Model 91

Mainframe?

Is this computer what you would call a “mainframe”? — HTGS (talk) 02:32, 30 October 2025 (UTC)

Different "you"s may have different answers.

The use of the term "mainframe computer" or just "mainframe" has evolved over time. It appears that "main frame" or "mainframe" originally largely referred to the main cabinet of electronics in a computer system - even a minicomputer system. Later, it was used for Big Iron computers.

The Model 91 was a specialized model of a line of computers the successors of which were considered to be "mainframes", even if the System/360 line wouldn't have been called that at the time. That could qualify it as a "mainframe".

It could also be considered "a type of computer with a high level of performance as compared to a general-purpose computer.", to quote the opening sentence of supercomputer, although it supported the IBM System/360 architecture, albeit with some incompatibilities and with decimal arithmetic instructions trapping to software implementations, so it could be considered a "general-purpose computer", albeit that would be massive overkill for computing the payroll. Guy Harris (talk) 10:26, 28 November 2025 (UTC)

Thank you. Do you think it’s worth adding either of those terms to the article at all? Presumably it could very well be couched in the sort of explainer you gave here.

I actually raise the question from a bit of a backward starting point: this infobox is used at the documentation for {{Infobox computing device}} (as you’re probably familiar) and I wanted to label it as an example of a mainframe (or whatever it is), so I can specify and add other examples of different device types. — HTGS (talk) 02:16, 4 December 2025 (UTC)

Do you think it’s worth adding either of those terms to the article at all? Given that almost any application for OS/360 would work, a 360/91 could run any application that a 360/65, for example, could run if, as per the Model 91 Functional Characteristics manual, there's software to trap and simulate the decimal arithmetic instructions - see page 62 of the Supervisor Services manual for OS/360 release 20.1, and if care is taken when writing assembler code to put in store and exception barriers or if care is taken to tell the compiler to generate barrier codes when necessary (see, for example, Appendix G of the PL/I Compiler (F) manual.

Given that, the Model 91 could, as per the cover of the Functional Characteristics manual, run "ultrahigh-speed, large-scale scientific and business applications." (emphasis mine), so it could function as a high-speed System/360, so it could be considered what is now called a "mainframe".

However, that would probably be overkill, as I think the Model 91 was very expensive (http://www.beagle-ears.com/lars/engineer/comphist/model360.htm says that "each of them was successively hand built by the same engineering team", as I suspected from the low production numbers), so I suspect that most if not all of them spent almost all their time running big number-crunching applications, i.e. operating as supercomputers. (I have some vague memory of reading that one of the supercomputer 360s/370s, either a Model 91/95 or a Model 195, as used for a transaction processing application, but that sort of conditional-branch-heavy application sounds like a waste of a supercomputer-microarchitecture S/360 or S/370.)

So, yeah, it is in some ways different from, say, a CDC 6600, as it's a supercomputer designed to run an instruction set designed for all 360 degrees of the computing range (yup, that's why it was "System/360"; see the compass rose used on some documentation), even though that may have been a bit of a waste. Perhaps the goal was to be able to reuse system software (I don't know whether any of the formatted-output code for FORTRAN and PL/I used the "edit" and "edit and mark" instructions, but, if they do, that might explain why the Model 91 hardware implemented it).

On the other hand, some stuff I found online seems to indicate that there was a COBOL compiler for the CDC 6600 (although I can't find a manual - bitsavers' 6600 stuff seems to have disappeared, and even the versions on the Wayback Machine don't seem to have 6600 COBOL documentation), so hey, what's a supercomputer from the 1960s other than "a mainframe that crunches floating-point numbers really fast"? :-)

this infobox is used at the documentation for {{Infobox computing device}} (as you’re probably familiar} Yes, I am. :-)

(I put that in to give something far from a laptop computer, as per the edit comment "Give a slightly less portable :-) example.", so people didn't think a "computing device" had to be a personal device you could carry.)

I guess you could add some current IBM Z machine if you want something that's more obviously a regular mainframe rather than a supercomputer with a mainframe instruction set, or something such as an IBM 3033 if you want something that's not subject to "that's a server, not a mainframe!" complaints.

@Chatul: @Peter Flass: @John Sauter: Any of you have anything to add here? Guy Harris (talk) 07:49, 4 December 2025 (UTC)

I would consider the 91 and 95 to be top of the line mainframes, and IBM intended them for the supercomputer market. As for the 6600 software, see . -- Shmuel (Seymour J.) Metz Username:Chatul (talk) 14:23, 4 December 2025 (UTC)

I would also consider the System/360 model 91 to be a top-of-the-line mainframe. My experience with small DEC computers (PDP-1, PDP-8, PDP-11) has led me to consider the large IBM, CDC and Burroughs computers (IBM 7090, 360/65, Burroughs B5500, CDC 6600) as mainframes. I don't consider the IBM 1401 a mainframe because it had a small footprint next to its contemporary, the IBM 7090. The IBM 360 line, like the DEC VAX line, had both mainframes and non-mainframes by my standards. John Sauter (talk) 21:12, 6 December 2025 (UTC)

References