Difference between revisions of "IBM 650"
m (Copyedit) |
m (+Furher reading) |
||
Line 14: | Line 14: | ||
A [[printer]] option was soon added; and eventually (September, 1956) also [[disk]]s, initially with a capacity of 6 million digits per drive, with up to four [[drives]]. | A [[printer]] option was soon added; and eventually (September, 1956) also [[disk]]s, initially with a capacity of 6 million digits per drive, with up to four [[drives]]. | ||
+ | |||
+ | ==Further reading== | ||
+ | |||
+ | * Charles J. Bashe, Lyle R. Johnson, John H. Palmer, Emerson W. Pugh, ''IBM's Early Computers'', MIT Press, Cambridge, 1986 |
Revision as of 15:02, 8 October 2018
The IBM 650 (formally the Type 650 Magnetic Drum Data-Processing Machine) was one of IBM's first computers. It was a very low-end design, using a drum for main memory, and was thus fairly slow, but it provided an entry into the computing world for those who needed more than a calculating punch (such as the IBM 604) but whose computing requirements did not justify the higher end machines such as the early IBM 700 series.
It was announced in July, 1953, and first shipment was in December, 1954; production continued until 1962. Due in part to its low cost, it was the first computer to have a production run of larger than the low ten's; eventually almost two thousand were produced (over 800 by the beginning of 1958 alone), and "it is safe to say that, for several years, far more computer scientists had their first experience on that machine than on any other".
Internally, it was decimal, rather than binary, using a bi-quinary representation; although inefficient in storage terms, this representation allowed extensive 'validity checking', since only 10 codes out of the 128 available (with 7 bits per digit) were valid. (The machine gained a reputation for reliability, due in part due to the validity checking.) The word length was 10 digits, although multiplication and division could use 20-digit values (for output and input, respectively).
It used a serial ALU (digit-serial, not bit-, though). The two internal registers in the CPU (a 10-digit 'distributor', and a 20-digit accumulator) used capacitor-based memory.
The instruction set was double-address, with a two-digit opcode, and 4 digits each of data address, and next instruction. (The fetch of the next instruction was overlapped with the execution of the previous one.)
The drum was 4" in diameter, 16" long, and ran at 12,500 RPM; with a complete rotation time of 4.8 msec, average access time was thus 2.4 msec (although optimum coding would reduce that for instructions). There were initially two models, with 10,000 and 20,000 digit memory capacity. Each 'band' (containing 5 tracks) contained 50 words; thus, 20 bands in the former, and 40 in the latter. The time to read a word, once it was at the heads, was only 100 μsec.
Initially, the only I/O was via punched cards, making it a small step for organizations whose data processing was already built around punched cards; the 650 could be used to simplify an existing punched card operation, and remove clerical steps. The reader punch could read 200 cards per minute, and punch 100, using 10-word buffers stored (but not addressable by the CPU) on the drum.
A printer option was soon added; and eventually (September, 1956) also disks, initially with a capacity of 6 million digits per drive, with up to four drives.
Further reading
- Charles J. Bashe, Lyle R. Johnson, John H. Palmer, Emerson W. Pugh, IBM's Early Computers, MIT Press, Cambridge, 1986