Help identifying a very old computer: 1974, maybe an NCR

Question

I am confident of most of the following information except oddly the manufacturer. I think that it was NCR but I am not quite sure. I have no idea of the model.

The machine had 32k of memory but 24 bit words not bytes. "Byte" was not yet in my vocabulary. Mostly, these words stored numbers: integers in 2's complement or floats. When characters were stored, they used a 6 bit code and were packed 4 to a word. Obviously, the coding was not ASCII, "ASCII" was also not yet in my vocabulary. There was no room for upper and lower case so it was usual to work entirely in upper case. If you insisted on lower case then you needed to use shift characters. Space was coded as 0. When I discovered ASCII later, I was surprised that space was 32, 0 had come to seem the natural code for space.

One reason that I am confident of this is that the number 2^23 = 8388608 is still burned into my brain as 8388607 was the maximum positive integer.

No hard disk and it was quite a while before I learned to use the tape drives.

The system accepted a few simple control commands. A typical program looked like this:

&JOB;3733;JOHN;MY FIRST PROGRAM
&OPER; PLEASE LOAD TAPE XXX123 H1
&FORTRAN;

Loads of ugly code with GOTOs and stuff

   STOP
   END
&RUN;

Data

&END;

FORTRAN was the most commonly used language but COBOL, ALGOL, and an assembly language called NEAT (I think) were also available.

So, can anyone confirm that it was a NCR or correct me?

Would anyone be able to guess the model?

Answer 1

Was it an Elliot-Automation 41xx system, an actual 24-bit contemporary of the 315 also sold by NCR during the mid-60s?

I came across one of these mentioned as a 32K word 24-bit system:

I started on an ICL 4120, which was upgraded to an Elliott 4130 (yes, the 40130 was made first, they were just at the time ICL for formed from Elliott and the others), the 4120 with rin core and the 4130 with plated wire storage (32K of 24 bit words!)

The Wikipedia page for Elliot Brothers (computer company) lists:

• Elliott 4100 series (1966) A joint development with NCR Corporation. Elliott selling to the scientific market and NCR selling to the commercial market.

Assorted references mention NEAT in association with this series of systems.

In fact, Barbosa (1977, 2016) had collaborated with the engineer Azevedo Machado in the coding of extensive source codes in FORTRAN, ALGOL, and NEAT in an NCR/Elliot 4130 [...]

(The Bloomsbury Handbook of Electronic Literature, 2017)

However, I've seen at least three different expansions for NEAT in association with NCR, so perhaps it shouldn't be taken as a reference to a single coherent language or technology.

It seems the 41xx line did not initially have a COBOL compiler, but eventually had one by the time it became part of ICL.

Answer 2

Definitely a 4130, I spent five years of my life maintaining these as well as coding in machine code. A 24 bit machine with a built in floating point unit and no chips only transistors running on a 10v supply. Happy days?!

Answer 3

Definitely looks like a 4100 series. I worked for NCR as a mainframe computer operator (NCR 315), then as a technical writer writing manuals for the NCR Elliott 4100 series, then in the Marylebone head office in Programming Information Services (with the unfortunate acronym of P.I.S.) where I was responsible for writing the operating system drivers that allowed the 4100 series to use NCR Card Random Access Memory devices ("CRAM").

Hope that helps. (And yeah, I'm still actively cutting code.) :)

Answer 4

Mostly off topic, posting to reminisce ...

In 1976, I think, I was hired because a bank's only computer programmer accepted a job offer elsewhere while in training for converting data from an NCR 315 to an NCR Century 101.

I'd never heard of an NCR 315, but had 3 hard years of experience on an NCR 100 and then a 101, running NCR's bank software package called C.I.F. (Central Information File). I was young, ambitious and a bit egotistical. At the end of my interview, the D.P. manager asked if I could do the changeover. "Sure, no problem".

As previously stated, the Century used 8 bit bytes (plus parity) while the 315 had slabs. When a slab contained two alphanumeric characters, it was easy to change them to two 8 bit bytes. But slabs containing 3 digits were a challenge. I gave up on writing my own routine and settled for a package utility called GDFC - General Data File Conversion. To test the program to convert 10,000 installment loans, I first had to run GDFC - which took an hour just to make the data useable to me.

I really enjoyed writing NEAT/3. When asked, I said it was like a fixed format COBOL. And, if needed, I could make the compiler shift to Level 2 - which was the equivalent of assembler. I later used that functionality to write an exception item pull program, which used information from reject reports to separate rejected checks so they did not have to be pulled manually.

Answer 5

Was it the NCR 315?

It used 12-bit slab memory structure using core memory. The instructions could use a memory slab as either two 6-bit alphanumeric characters or as three 4-bit BCD characters. Basic memory was 5000 "slabs" of handmade core memory, which was expandable to a maximum of 40,000 slabs in four refrigerator-size cabinets.

...

A numeric value contains up to eight slabs.

So it supports numeric operations on words in multiples of 12 bits, up to 96 bits.

Available Languages

• NCR Assembler Language

• National Electronic Autocoding Technique (NEAT)

• COBOL

• BEST

Hmm, Fortran and Algol are missing from that list. Maybe you are thinking of a successor that inherited the NCR 315's 12-bit words?

Edit: Sorry, I didn't read the comments before posting this or I would have seen the suggestion that it might be the 315 or the 615. I guess great minds think alike!

Answer 6

I programmed NCR Century computers (the 615 series) from 1973. These were 8 bit bytes, in 16 byte words. Each byte held 1 Ascii character, or 2 digits 0-9. "Byte" was invented in 1956 (when I was 7!). Burroughs B3500's used 8 bit bytes and Ascii, as did IBM, who used EBCDIC code. Holding text in 6 bits, in 24 bit words was how the ICL 1900 series worked.

Answer 7

This is a reach, but how about an SDS 930?

It had 24 bit words and ALGOL. Don't know about the rest of it.