A compact bench machine in a chip works in California under fluorescent light, a two-tiered cream-white housing with rounded edges and a dark base strip, on top a brushed aluminium plate, across it a dark grey-blue carriage with two knurled chrome knobs, in its middle a round turntable with a toothed steel ring holding a mirror-bright blue-grey silicon disc five centimetres across covered with a delicate grid of hair-thin silvery lines, above it a knurled chrome head with a fine diamond point, a cream arm rising to a microscope eyepiece, a sheet of tissue paper, a glass dish of tiny broken square chips.
What it was used for
The machine scribed lines on the silicon wafer so the chips could be separated before packaging; the Computer History Museum holds such a unit by Tempress Research of Sunnyvale, about 1965, with lens attachment, eyepiece, die platform and dust cover, weighing 41 pounds. A documented user is NASA's Electronics Research Center in Cambridge, Massachusetts, whose 1970 report states that the Tempress scribing machine was used in all their experiments. Breaking was then done by hand, with a hand-held roller of 1 inch diameter. The wafers tested were 8 to 10 mil thick and separated into dice of 0.061 by 0.068 inch.
What was new
Scribing itself was not new in 1965: the NASA report calls the diamond-tip method the industry standard for many years and lists 4 methods in use or under test at the time - diamond cutting wheel, diamond point, laser and chemical etching. The maker's 1966 advertisement describes the machine as having been in development for over 2 years, so as a new generation rather than an invention. What characterised it was that the machine secured the tool's angle and pressure; the maker advertises exactly that, though in a conditional clause. The barb lies in the factory setting: at the normal 36 strokes per minute, tool bounce at the wafer edge was virtually impossible to eliminate, with excessive peripheral losses; only halving it to 18 produced acceptable lines.
What the sources leave open: No price for the machine is documented; the advertisement speaks only of a surprisingly low price. Production or sales figures are missing. A 1965 Wescon industrial design award rests on a single Computer History Museum page and is not independently confirmed. The NASA measurements apply to oxide-free (111) silicon.