A survey of room-sized computers

Some great photos and links in the answers to the question How big is a mainframe? over on the Retrocomputing stackexchange - here are just a couple:

(Strela computer)

(Shuttle Mission Simulator)

This post prompted by a nearby comment:

which reminded me of the link and provokes the thought that building a home around a computer, especially a large computer, could be quite appropriate. One can even make a home inside a computer:

DEUCE had a door at the back of the cabinet-frame and one went inside to change the valves or to ‘beat them to death’ or whatever. It was also quite nice and warm and fairly private. I have to confess that I did some of my courting inside these machines and, on one occasion, after I had married, I went late one night to fix a machine at Warton, near Blackpool: I took my wife with me. The Machine Room itself was cold for the reasons I’ve given but it was warm inside the computer so she took a book and a chair inside!

(P. J. Walker helpfully provides an illustration with this remark:


There’s an image of a pair of DEUCE (and much besides) in this timeline:


This is a “room size computer” I helped build in the early 90’s:


It’s a Meiko CS2.

This was Livermore Labs in Ca. in the early 90’s. It was, for a short time, the fastest supercomputer on the planet.

To give an example of scale - it’s 2 meters high. Each individual small box has 4 CPU cards in it, although the middle set of 4 boxes have a part of the switch network. (the cables run diagonally under the floor - shortest route) The small 4-box wide unit in the middle has the top of the switch network and a set of hot spare CPU cards, 4 general user CPU cards and 8 filestore CPU cards. The black boxes are RAID units (3 arrays of 16GB each) which were eventually moved into the system itself.

Specifications? 256 ‘nodes’ of a dual-sparc running at 88Mhz initially, faster once it was finally commissioned. Each node has 128MB of very fast (for the time) DRAM with interleaved access and a local 1GB SCSI drive. The 16 hot spares were the same, as were the other IO and console nodes. The thin dark line in the middle top and bottom is LEDs, 16 per node because everyone likes flashing lights. Every node ran a Solaris with patches and drivers for the data and system control networks. (We had power control and remote console access to every element as well as some switch monitoring, etc.)

I don’t recall the overall power budget for it (a few 10’s of Kw maybe), but it was all air cooled and loud. Probably why I have bad hearing now as I spent months in there.

Today I can hold more pure CPU power and storage with a box on Raspberry Pi’s, but back then…




I remember in the 90’s installing a computer in a server room about 1/2 the size of a school hall… My role was to put a Novell Netware system in so they can switch off the Mainframe. It was Novell ELS Level 1 in those days… My CNE course counted for something.

I was amused at the size of the Novell server compared to the mainframe

1 Like

I worked at a shop in the late 90s that had a decommissioned IBM mainframe in the server room that had been shut off and still had the last core dump hanging out of the line printer. It was replaced by a (probably Pentium Pro, at the time?) Compaq server running NetWare. When they pulled out the mainframe, we had a cavernous raised-floor computer room with viewing windows and massive air ducts and all the trappings … with two full tower cases sitting on a lone Steelcase desk at the end: the server and a high capacity tape robot.


The university computing center I worked for in the 1990s had an IBM 3090 (pic: https://www.ithistory.org/db/hardware/ibm/ibm-3090-processor-complex) which they finally got rid of (some of the university administration software had been running in it, past tense). The subcontracting hauling company was obviously very specialized: the computer technicians and the moving men had a fascinating choreography. As the route they had to take from the raised-floor machine room to the trucks waiting outside was quite contorted, they ended up leaving pieces of the machine sprinkled alongside the hallways, waiting for the next stage of dismantling and packing.

I had nothing to do with the move but hung around, and overheard some of the students wondering about the cabinets: “What are these things?” Mind, these were man-height steel frames, with copper piping for the water cooling. Some of the smarter students noticed some wiring and boards and opined “Hey, these are maybe computers?” To which the other students: “NO WAY!”


It’s a funny thing, what people think of when they think of a computer. Partly generational of course. I think the biggest computer I’ve seen might be the Pegasus by Ferranti, in its final month on display at the Science Museum before going into storage:
Google Photos
(More photos from that retro-meetup in this album)


I’ve replaced my share of behemoth mainframes and “washer/dryer” sized IBMs with Tower computers or pizza boxes. I’ll never forget the size of the power cable these things used.

But it should be noted, rumor has it that back in the day, Novell used their server room as a source of heat for their building.

Not exactly room sized, yet these two are one of the first Swedish build and Danish build computers.

BESK (Sweden 1953) and DASK (Denmark 1955)

I know they are not much compared to American, English and so on. Yet they were revolutionaery in Scandinavia at that time.

1 Like

Both very interesting machines, I’d say! DASK looks room-sized in this video:

DASK in context of Scandinavian computing history in this talk:

This is the only model from Regnecentralen, that I have ever used in person. Not room sized at all.

Piccoline 80186 Based School computer

I did hear about Dask being a room sized at school, yet I have never seen a picture of the complete installation of that machine.

MIT’s TX-2 was often described as a “computer with a roof” – so it would have been the room quite literally.

(Which provides a bridge to yet another category, computers with a room in it. E.g., the UNIVAC I, there are stories of operators actually having placed a small coffee table and a chair inside.)

1 Like

Ah yes - UNIVAC I - the canonical example! The main part of the machine, called the frame, was an 8’ by 10’ box, with a ceiling. Photos here and here.

Resources via this video:

An educational overview of UNIVAC 1 basic components. Intro segment includes higher quality images than are contained in the original film. Original 1953 film begins at index 5:52, followed by brief photo gallery.


Some room sized computers came furnished:


Considering how long you had to wait for even room size computers to do anything, comfortable seating is desirable.

I came across a relevant article here:
Programming the ENIAC: an example of why computer history is hard

ENIAC – the Electronic Numerical Integrator and Computer – was a room-sized machine with over 17,000 vacuum tubes. It started running at the end of 1945, and for five years it was the only fully electronic computer running in the US. Estimates are that by the time it was retired in 1955, it had done more calculations than all human beings in all of history.

It argues that ENIAC was put to practical use right from the beginning with a substantial program: “an 840 instruction program that used a subroutine, nested loops, and indirect addressing for both data locations and jump destinations” and a useful one, being a simulation of nuclear fission which “broke new ground by using a digital computer as a real-world simulator for problems that cannot be solved using traditional mathematics.”

It’s a short piece hanging off a book “ENIAC in action”

1 Like

The Manchester Baby was room-sized, as seen here with a young Dai Edwards in attendance:

That page links to a number of short clips of oral histories collected by The British Library - with transcipts and links to full interviews. There are more of those to be found: hardware, and software (with considerable overlap).

A slightly less young Dai Edwards gave this talk in 1981:


In 1949, Professor Dai Edwards of Manchester University began working with Tom Kilburn on the prototype of the Manchester Mark I. The Mark I was arguably the first stored program computer, put into service in April 1949. Professor Edwards covers the evolution of computing devices and technologies of the late 1940s and early 50s. Edwards worked closely with F.C. Williams, developer of the Williams Tube, and he details Williams’ work at the University and also points out other significant contributors to Manchester’s computer program, including Alan Turing. He discusses Manchester’s “problem of memory” and describes the use of the Williams Tube. Edwards outlines the design of the early Manchester computers, starting with the “Baby” machine, followed by the Manchester Mark I, and then the Ferranti commercial computers based off the Manchester University designs. He goes through the methods of programming the Manchester computers and the problems using the machines. The lecture concludes with questions from the audience.

Somewhat related, here’s another talk about the early Manchester machines, this time by Prof. Richard Grimsdale:


Professor Richard Grimsdale, then of University of Sussex, previously of Manchester University, is often credited as the creator of the earliest transistorized computer. Prof. Grimsdale describes his work at Manchester, including his work with with Tom Kilburn, Dai Edwards, and Freddie Williams, and how the individual circuits of the Manchester transistorized computer operated, as well as storage techniques for magnetic drum and cathode ray tube systems (Williams-Kilburn Tube.) Professor Grimsdale describes the evolution of the machine, including the experimental core memory and issues with individual transistors used in the project as well as the use of Grimsdale’s design as the basis for the Metrovick 950. He ends his prepared remarks with a description of his work on the Atlas computer, and the rod memory the machine used. Audience questions follow, including discussion of Alan Turing’s work with Grimsdale.