The IBM 7070 was IBM's first commercial transistor mainframe; the first announced (in September, 1958), as the IBM 7090 beat it to actual first delivery. It was intended for commercial use (as it was built at a time when computers for scientific and business computing used separate instruction sets).
It was built out of Standard Modular System cards; the complete machine contained about 14,000 of these cards. It mostly used Complemented Transistor Diode Logic (CTDL), a form of DTL. Main memory was core, either 5K or 10K words.
Like its vacuum tube technology predecessors, the IBM 650 and IBM 705, for which it was the intended replacement, it was a decimal machine; words were 10 digits long, plus a sign digit. Internally it used a two-out-of-five code, unlike the 650, which had used bi-quinary coding; with the higher per-bit cost of core memory, a coding which used less bits per digit made sense.
It was not upwardly compatible with its predecessors, however. There was good reason for the 650 incompatability; that machine used a drum for main memory, and the instruction format included a 'next address' field, which was not needed with the random access core memory of the 7070. The incompatability with the 705 was harder to justify, and in fact IBM was forced to bring out a compatible 7000 series replacement for the 705, the IBM 7080.
Up to twelve magnetic tape drives could be connected to a 7070, via two separate data channels; the 729II could read 15,000 characters per second, the 729IV could read 62,500. These drives had a read head mounted just after the write head in the tape path, to allow immediate checking while writing.
The 7300 moveable-head disk system could store up to to 6 million digits per drive; up to four drives could be connected, via the same two data channels. Each disk had three sets of heads, to minimize access time.
- Charles J. Bashe, Lyle R. Johnson, John H. Palmer, Emerson W. Pugh, IBM's Early Computers, MIT Press, Cambridge, 1986