A basement workbench in a house in New Jersey on an autumn evening, a half-built home computer on an orange-brown perforated board twenty centimetres long, in its centre a white ceramic package five centimetres long with forty silver legs and a small grey metal lid, plain smooth and unmarked, around it black plastic chips, a small gold crystal can, in front a bare aluminium plate with a row of eight black toggle switches, a young hand in a plaid flannel sleeve flipping one switch, two tiny red display windows glowing as a soft blur, a soldering iron in its stand, a pegboard of tools on the wall.
What it was used for
A build-it-yourself computer from Popular Electronics, on the cover of the August 1976 issue, written by Joseph Weisbecker, an engineer at RCA Labs in Princeton. The Sarnoff Collection gives about 80 USD for the parts, about 100 USD with a video display, against nearly 400 USD to build an Altair 8800. Input was through 8 toggle switches, with a hexadecimal LED display, 256 bytes of RAM and few chips, notably RCA's CDP1802. Users wrote simple programs, played games and monitored intruder alarms. Later articles brought a cheaper display, a hex keyboard, a 9-volt battery to retain memory contents, and the CDP1861 video chip, which displayed a 256-byte segment of memory on a television.
What was new
What the chip could do: work in CMOS and therefore draw very little power. IEEE Spectrum notes a radiation-resistant version, which is why 6 of them ran the Jupiter probe Galileo. It was not without forerunners: per the same source the 1801U and 1801R appeared in 1975, and combining them produced the 1802 in 1976. For the Elf the decisive thing was price: the Sarnoff Collection gives minimising system cost as Weisbecker's primary goal. That this happened in a house which in 1971 had sold its computer division at a 490 million USD loss and largely ended computer research is the contradiction in the machine - per the same source the COSMAC VIP and Studio II (1977) were both commercial failures.
What the sources leave open: The Sarnoff Collection page contradicts itself: in one place it gives the August 1976 cover, in another a debut in the August 1974 issue. IEEE Spectrum gives August 1976, which is followed here. Both pages give about 80 USD. Neither gives a figure for how many were built. IEEE Spectrum calls the 1802 the first CMOS processor, but Intersil was already making its CMOS 6100 processor in 1975.