In search of 48 bit computers

I’ve been hunting for 48 bit computers - and found a few. Let me know of any I’ve missed…

I found a few in “A Third Survey of Domestic Electronic Digital Computing Systems” a huge 1961 document from the Ballistic Research Laboratories. From there, for starters, here’s the transistor-based Philco 2000:

Here’s a video about Philco and their machines:

Now let’s see what’s in the Third Survey that merits a 48 bit description - detailed technical descriptions and some photos and diagrams at the links:

  • SAC Data Processing Subsystem AN/FSQ 31 (V) from IBM with 48 bit words plus 2 parity bits and offering 6 bit bytes
  • AN/FSQ 32 again from IBM… same model? Certainly very similar
  • Control Data Corporation Model 1604 a 48 bit word with means to access 24 bit subwords
  • Datamatic 1000 Electronic Data Processing System has a 52 bit word with 48 data bits and 4 check bits
  • Honeywell 800 has 48 plus 6 check bits. “Designed for general purpose business, business-scientific, and scientific applications, system capable of running eight programs simultaneously.”
  • MANIAC II (Mathematical Analyzer Numerical Integrator and Computer Model II) is tube-based, has 4k words of core storage and 12k of CRT storage. “Average error-free running period: Several Hours”
  • MANIAC III is transistor-based, more than twice the performance.
  • MERLIN just over 1000 tubes, 8k words of electrostatic barrier grid tube storage
  • NAREC Naval Research Electronic Computer uses 1300 tubes, has 16k words of core, has 48 bits and packs two instructions per word. Needs 31 staff to offer two-shift access, at about 85% availability.
  • PHILCO 2000
    Philco Transistorized Automatic Computer (TRANSAC S-2000) as seen in video above. 20k transistors. 4 microsecond add time (variable.) “There is one 48 bit register, three 24 bit registers, and up to 32 optional index registers. The program section has asynchronous logic. Automatic assembling and compiling system called TAC”
  • PHILCO CXPQ Philco Transistorized Automatic Computer CXPQ. 5500 tyransistors, 4k words core, 16k words drum, 45 microsecond add time, 7 index registers.

From 1961, we have Burroughs B5000 Computer and there’s a 1979 account STORIES ABOUT THE B5000 AND PEOPLE WHO WERE THERE

The B5000 was designed to be an Algol machine.

So the worlds greatest computer expert returned to Europe and told his friends about the amazing B5000 Algol system. They decided to order three B5000s.

Wikipedia says:

The Burroughs Large Systems Group designed large mainframes using stack machine instruction sets with dense syllables and 48-bit data words. The first such design is the B5000 in 1961. It is optimized for running ALGOL 60 extremely well, using simple compilers. It evolved into the B5500.

And from West Germany, from 1964 to 1971, we have “the Telefunken TR440, a machine with a 48-bit word length”

I also found some info on the use of 48 bits as a floating point data size:

The historic English Electric KDF9 computer used a floating-point format very similar to that of the IBM 7090 computer, except for being adapted to its 48-bit word length.

One solution would be to use 48-bit integers, and 48-bit and 96-bit floating-point numbers, as the ICL 1900 did

The Scientific Controls Corporation 660 computer also used, as its normal floating-point format, the full 48-bit format shown in this section of the diagram.

via @oldben’s enquiry about the fate of surplus “48+ bit” mainframes


Ferranti Orion?

“The basic Orion machine included 4,096 48-bit words of slow, 12μs, core memory, which could be expanded to 16,384 words.[10] Each word could be organized as eight 6-bit characters, a single 48-bit binary number, or a single floating-point number with a 40-bit fraction and an 8-bit exponent. The system included built-in capabilities for working with Pound sterling before decimalization.[13] The core memory was backed by one or two magnetic drums with 16k words each.[14] Various offline input/output included magnetic disks, tape drives, punched cards, punched tape and printers.”

1 Like

I’ve an obscure one: The “Mailüfterl” (built 1956-1958) had a 48-bit word length.

Instruction layout (“Freiburg code”):
4 bits (1 bit used) flag binary/decimal
4 bits conditional execution codes (15 conditions)
4 bits main instruction code (15 codes)
3 bits auxilary instruction code (7 codes)
9 functional bits (e.g., flags for storage select)
2 × 4 bits address index (10^2)
4 × 4 bits address (10^4)

I guess, there may be other early one-of-a-kind machines using this kind of mirco-level instruction layout, especially relay computers.

Edit: More details here, Norbert Kehrer’s Mailüfterl emulator page.


Atlas? A joint development effort.

  • 48-bit word size. A word could hold one floating-point number, one instruction, two 24-bit addresses or signed integers, or eight 6-bit characters.

The Atlas Computer was one of the world’s first supercomputers, in use from 1962 until 1971. It was considered to be the most powerful computer in the world at that time.[1] Atlas’ capacity promoted the saying that when it went offline, half of the United Kingdom’s computer capacity was lost.[2] It is notable for being the first machine with virtual memory (at that time referred to as ‘one-level store’) using pagingtechniques; this approach quickly spread, and is now ubiquitous.


It’s interesting to contrast the old architectures that were, apparently, much more focused on numeric computation than simply “data processing”.

Back on the CDC systems, with their 60 bit word, and 6 bit character set. 10 character per word, lowercase took 12 bits since you prefixed a normal character with a ^.

Simply, if you were doing TEXT processing, you had to jump through hoops (and thus had efficiency costs) to do the processing in contrast to numeric works that could use the native word size.

Of course, today, while we’re back up to 64b architectures and large register spaces, the machines are more adept at working with the smaller word sizes (8 and 16bit today for text processing). Arguably, the vast majority of modern data processing is more text based than numeric, at least in terms of actual data moved. Arithmetic (both floating and fixed point) is still very important, but now it all comes with a boatload of extra data to track with it.

Mind, I’m making all this up.

I love it. :smiley: I guess before crypto instructions, you had to get your competitive edge somewhere.

The IBM ACS-1 project (1961-1969) was a 48 bit machine that first implemented a bunch of features now seen in modern processors.

Marvellous! and thanks @B.B and @NoLand for your findings too. Some great machines.

1 Like

A couple more for your collection:-
Control Data Corporation 3400 and 3600
FACOM 212 (a word was 12 base-10 digits long)
BESM-6 (Soviet machine)
109C, DJS-6, DJS-8, Model 111 (Chinese, see Pg 20 & 23)
Telefunken TR440 & TR4 (with a mouse?!)

1 Like

The ICT 1301.
From Wikipedia

The 1301 was designed by an ICT and GEC joint subsidiary, Computer Developments Limited (CDL) at GEC’s Coventry site.[1]

The 1301 was the main machine in the line. Its main memory came in increments of 400 words of 48 bits (12 decimal digits or 12 four-bit binary values, 0-15) plus two parity bits. The maximum size was 4,000 words. It was the first ICT machine to use core memory.

The computer was announced in May 1960. The first customer delivery was in 1962, a 1301 sold to the University of London.[2] About 150 to 200 of these machines were sold. One of their main attractions was that they performed British currency calculations (pounds, shillings and pence) in hardware. They also had the advantage of programmers not having to learn binary or octal arithmetic as the instruction set was pure decimal and the arithmetic unit had no binary mode, only decimal or pounds, shillings and pence.

ICT 1301 machine serial #6 “Flossie” had an interesting history. The first to be installed at a customer site at the University Of London, there were five previous machines, but they were used for software development and design troubleshooting.

The ICT 1301 Restoration Project

It was scrapped by the University and bought by a group of students who restored it and ran it as an accounting bureau for five years, scrapped again and given away for free via the Amateur Computer Club magazine, stored in a barn for 25 years and then restored again with the help of the Computer Conservation Society, was again at risk of being scrapped and now in storage at The National Museum of Computing at Bletchley Park, England.