Re:The Electronic Computers, Part 2: Colossus
by Animats on Saturday September 23, 2017 @01:20PM
(#15320705)
Good article. It gives Tommy Flowers the credit. He doesn't get mentioned much.
Colossus wasn't a computer as we think of one today. It was a key-tester, like a Bitcoin ASIC. No memory, just a few registers. No stored program.
What we think of as a computer today is a general purpose, stored program, digital machine. Colossus was a special-purpose, hard-wired program digital machine. There were many such special-purpose machines before full-fledged computers appeared.
Just before WWII, several groups were working on electronic arithmetic units. A. Halsey Dickinson at IBM was working on vacuum tube arithmetic, and by 1942, had an electronic multiplier.[1] Some of those circuits were used in the ENIAC, and by 1946, IBM had an electronic multiplier as a product, the IBM 603.
IBM's 600 series machines somehow don't seem to be mentioned much in the history of computing. They were commercial products, so they had to work reliably and be cost-effective. The IBM 601 (1936) was the first fully automatic multiplier - electomechanical, punch cards in and out, a few registers of memory. The 602 and 602A added division. The 603 did the same job as the 602, but using vacuum tubes. The 604 (1948) was plugboard programmable with up to 40 program steps. There were more 600 series machines, up to the IBM 650, on which Knuth learned to program. The 650 had a magnetic drum and real stored programs.
The big problem in the early days was memory. It was clear by the mid-1930s that electronic arithmetic was possible, but nobody had a good storage device yet. The early approaches were recirculating delay lines (slow, sequential access, not much capacity), magnetic drums (slower, sequential access, more capacity), and Williams tubes (faster, expensive, not much capacity, but random access). Core memory was the first technology that really worked well. IBM developed automatic looms for weaving it, which made it sort of affordable. Still, through the 1960s, memory was about a million dollars a megabyte.
There was a whole sideline of special-purpose electronic computing devices in the 1950s and 1960s, when general-purpose computers were too expensive. Teleregister did special-purposes machines for stock markets and airline reservations. American Totalizator did special purpose machines for racetracks, and was an investor in Univac. There were cute little machines such as the Mathatron (an early calculator the size of a desk) and Mem-O-Ree (an inventory control system). These all had electronic computing components but were not full general purpose computers.
Colossus wasn't a computer as we think of one today. It was a key-tester, like a Bitcoin ASIC. No memory, just a few registers. No stored program.
What we think of as a computer today is a general purpose, stored program, digital machine. Colossus was a special-purpose, hard-wired program digital machine. There were many such special-purpose machines before full-fledged computers appeared.
Just before WWII, several groups were working on electronic arithmetic units. A. Halsey Dickinson at IBM was working on vacuum tube arithmetic, and by 1942, had an electronic multiplier.[1] Some of those circuits were used in the ENIAC, and by 1946, IBM had an electronic multiplier as a product, the IBM 603.
IBM's 600 series machines somehow don't seem to be mentioned much in the history of computing. They were commercial products, so they had to work reliably and be cost-effective. The IBM 601 (1936) was the first fully automatic multiplier - electomechanical, punch cards in and out, a few registers of memory. The 602 and 602A added division. The 603 did the same job as the 602, but using vacuum tubes. The 604 (1948) was plugboard programmable with up to 40 program steps. There were more 600 series machines, up to the IBM 650, on which Knuth learned to program. The 650 had a magnetic drum and real stored programs.
The big problem in the early days was memory. It was clear by the mid-1930s that electronic arithmetic was possible, but nobody had a good storage device yet. The early approaches were recirculating delay lines (slow, sequential access, not much capacity), magnetic drums (slower, sequential access, more capacity), and Williams tubes (faster, expensive, not much capacity, but random access). Core memory was the first technology that really worked well. IBM developed automatic looms for weaving it, which made it sort of affordable. Still, through the 1960s, memory was about a million dollars a megabyte.
There was a whole sideline of special-purpose electronic computing devices in the 1950s and 1960s, when general-purpose computers were too expensive. Teleregister did special-purposes machines for stock markets and airline reservations. American Totalizator did special purpose machines for racetracks, and was an investor in Univac. There were cute little machines such as the Mathatron (an early calculator the size of a desk) and Mem-O-Ree (an inventory control system). These all had electronic computing components but were not full general purpose computers.
[1] http://www-03.ibm.com/ibm/history/ibm100/us/en/icons/ibm603/...