(Digital Extremities home page)

Computer Hardware Basics

"How can something that seems so physical only be tiny signals scribbled into chips, that disappear once the PC is turned off?"
– Brenda Lobb (AAHS '08)

Chapter: Computer Hardware Basics

(back)

Relays and Vacuum Tubes: WW II and the First Generation

Relay
Relay (Wikimedia photo)

At the time of WW II, transistors and computer chips didn't exist, and the primary technologies available for making computer logic circuits were relays and vacuum tubes.

Relays are mechanical switches, controlled by electromagnets. They are very reliable, cheap, and use little power, but are very slow compared to electronic devices. Relays were invented in 1835.

The first computer bug was literally a moth, stuck in a relay of the Harvard Mark II, that kept it from functioning. Below is the page from the computer's log from September 9, 1947, with the famous moth:

The First Computer Bug: A Moth
The First Computer Bug: A Moth (Wikimedia photo)
Vacuum tube
Vacuum Tube
(Wikimedia photo)

Vacuum tubes are electronic devices that can also be used to switch electrical currents on and off. They differ from relays in that they have no moving parts, so they can turn on and off about 1000 times faster. But vacuum tubes require an electrical filament (a very hot wire, like in an incandescent light bulb) to operate, so they use lots of energy, generate lots of heat, and can burn out. Vacuum tubes originated in the early 1900s.

World War II Computers

World War 2 created a demand for computing machines to speed up the solution of some urgent problems, in particular ballistics and code breaking.

Ballistics is concerned with the problem of shooting artillery shells so that they land where you want them to land. This isn’t a trivial bit of arithmetic, and if your calculations are incorrect or not sufficiently precise, your weapons are useless, or worse than useless. Gonick cites a case in World War 1 when " … the German canon 'Big Bertha' shot 94 miles – twice as far as expected from oversimplified calculations … " Oops.

Two important early computers in the US were developed for ballistics calculations: the Mark I and ENIAC. A third early computer called Collosus was developed in Britain, and was concerned with codebreaking. (There are several other early computers of note, especially the Atanasoff-Berry Computer (or ABC) at Iowa State, and the Z3 built by Konrad Zuse in Germany. The ABC is considered to be the world's first electronic digital computer, but it was not intended to solve one particular kind of math problem (systems of linear equations), and was not programmable.)

Mark I

Mark I
Harvard/IBM Mark I (photo © IBM)

The Mark I was a joint project of Harvard University and IBM, conducted for the Navy. It was an enormous electro-mechanical calculator, over 50 feet long, 8 feet high, and 2 feet wide, and was directly inspired by Babbage's Analytical Engine. It used thousands of relays, and was controlled by punched paper tape. It had no conditional jump instruction, and a loop was accomplished by attaching the end of the tape to the beginning, literally forming a paper loop. It could multiply two 23 digit numbers in 5.7 seconds, and divide them in 15.3 seconds (link) – certainly better than a person, but glacial by any modern standard. Although it was slow, its use of relays rather than tubes made it extremely reliable &ndahs; and is said to have also made it sound like a thousand knitting needles when working. It was first operational in May of 1944.

ENIAC

ENIAC
ENIAC (U.S. Army photo)

ENIAC
ENIAC (U.S. Army photo)

ENIAC (for Electronic Numerical Integrator and Calculator) was another computer intended for ballistics calculations, this time for the Army. It was built at the University of Pennsylvania, and it used vacuum tubes rather than relays. It had some serious limitations, but it was fast, about 1000 times faster than anything that had come before. It required an enormous amount of electricity, it gave off a lot of heat and required its own air conditioning system, and it could only be programmed by moving patch cords and changing the settings on switches. It had 17480 tubes, and failures were common. Tubes frequently failed when the machine was turned on or off, but by reducing power levels and leaving the machine on most of the time they were able to consistently operate for more than 12 hours at a time before encountering a failure – not great reliability by current standards, but good enough to do useful work, and better than had been predicted. ENIAC could do about 5000 ten-digit addition or subtraction operations per second. Although ENIAC was completed too late to contribute to the outcome of WWII, it was used to help design the hydrogen bomb, predict the weather, and design wind tunnels, among other things. It became operational in July of 1946, and its ten years of operation, it is estimated to have done more arithmetic than the entire human race had done up to 1945.

Colossus

Bletchley Park
Bletchley Park Copyright Bletchley Park Trust

Another extremely important computer of this era was the Colossus, actually a series of computers built by the British at Bletchley Park and used to decode intercepted German messages. The first one was put into service by February of 1944, and ten were in use by war's end. It is regarded by many as the world's first electronic digital computer.

Colossus
Colossus

The Colossi were special-purpose machines, with limited programmability, but they were extremely fast and effective, and made an important contribution to the victory of the Allies. Among other things, they allowed Allied commanders to know that their campaign of deception leading up to D-Day had been successful. The existence of Colossus was kept secret for many years after the end of WW II, and so it is not as well known as the ENIAC and Mark I.

UNIVAC 1

Univac 1
UNIVAC 1 Photo Courtesy of Hagley Museum and Library

The UNIVAC-1 was the first mass-produced computer in the US, and was designed for business rather than scientific applications. The first sale was to the US Census bureau, in 1951. J. Presper Eckert and John Mauchly, who had been the primary designers of the ENIAC, did much of the design work. Like ENIAC, it used vacuum tube technology (5200 tubes), and like ENIAC, it used a large amount of power (125 kilowatts). It could do about 1900 additions or 465 multiplications per second. 46 of the computers were eventually sold.

A program running on a UNIVAC 1 was famously used by CBS to successfully predict the outcome of the 1952 presidential election. According to George Gray, "This was enormously favorable publicity for the UNIVAC I, and for several years the name UNIVAC was synonymous with computer in the public mind. . . This public confusion of UNIVAC with computer was extremely galling to the leadership of IBM." (link)

SAGE

SAGE (for Semi-Automatic Ground Environment) was another notable computer system from the 1950s, built with vacuum tube technology. SAGE was an immense U.S. Air Force cold-war project, designed to detect and defend the U.S. against Soviet bomber attacks. It was not a single computer, but a network of 24 ENORMOUS computers, connected to each other and over 100 radar stations around the country via telephone lines. Each computer weighed 250 tons, and required a 3,000kW power supply (enough to operate 30,000 100 watt light bulbs) and contained over 49,000 vacuum tubes. These were the largest computers ever built. For reliability, each site contained two of these computers, with one active and one running on "hot standby" in case the active one failed.

SAGE Building at Mcguire AFB
SAGE building at McGuire Air Force Base, NJ.
Picture used with the permission of The MITRE Corporation.
Copyright © The MITRE Corporation. All Rights Reserved.

SAGE Building at Mcguire AFB
Floor plan of SAGE Direction Center
Picture used with the permission of The MITRE Corporation.
Copyright © The MITRE Corporation. All Rights Reserved.

The SAGE system was designed to do real-time processing, in which the system has to process input data and respond quickly enough to be useful. In this particular case, it had to process input from the radar sources, compare it against information about scheduled friendly aircraft flights (such as commercial airlines) to detect suspicious activity, and report the results quickly so that if necessary, defensive missiles and aircraft could be deployed to the correct locations before the attackers reached their targets. (Contrast this with batch processing, such as printing out payroll checks or running monthly reports, where the timeliness is much less critical – generally speaking, your hometown won't be incinerated because your paycheck was printed at 12:05 instead of 12:03.)

SAGE Computer
SAGE computer frames
Picture used with the permission of The MITRE Corporation.
Copyright © The MITRE Corporation. All Rights Reserved.

Each SAGE site could support up to 150 operators, each with a graphical display tube and a "light gun" pointing device. This was incredibly advanced – remember, this is taking place in the 1950, in the punch-card era. Each display station had a built-in ashtray and cigarette lighter, as well.

SAGE Display Tube
SAGE display tube and light gun
Picture used with the permission of The MITRE Corporation.
Copyright © The MITRE Corporation. All Rights Reserved.

It was an enormous development effort, with over 7000 programmers involved. It was a joint project of IBM, Western Electric, MIT/Lincoln Laboratories, Burroughs, The RAND Corporation, and a host of smaller companies. At one point it utilized of 20% of all the programmers in the world. The estimated cost of the system was between $8 and $12 billion in 1964 dollars. It was operational from the late 50s through the early 80s, but was essentially obsolete before it was completed – it was not fast enough to respond to the new intercontinental ballistic missiles (ICBMs).

Next: computing with transistors and chips.