An incoming inspection bench at a computer maker in California, on a grey antistatic tabletop rows of identical dark ceramic chips in strips of black foam, all plain smooth and unmarked, in the middle one chip in a test socket under a ring lamp, a dark purple-brown ceramic body nineteen millimetres long with eight gold-plated slightly tarnished legs on each side, its lid cut away, a gold frame around the open cavity, inside a silvery rectangular die divided into four bright fields by a central strip, fine bond wires at both short ends, a hand holding tweezers above it, a grey test cabinet with plain smooth panels soft behind.
What it was used for
Computer makers electrically retested delivered memory devices before they went onto boards. What such an inspection could turn up was shown by Richard W. Anderson of Hewlett-Packard at an EIA conference on 24 March 1980: according to the editor John Minck's foreword to Anderson's memoir, a study of over 300,000 DRAMs from American and Japanese suppliers in which Japanese parts met incoming specifications "up to 5 times better than the US" and were "up to 10 times more reliable" in the field. The 1990 OTA study gives 16K DRAM failure rates in its Table C-1, in units it labels "parts per million": 0.24 for Japanese against 1.00 for US vendors in 1978, and 0.05 against 0.02 in 1982.
What was new
With the 256K device of 1985 what was being looked for shifted. The object of official examination was no longer the quality of the delivery but its price: on 11 December 1985 the US Department of Commerce self-initiated - with no company petition - an antidumping investigation into DRAMs of 256K and above from Japan and estimated the dumping margin for June to October 1985 at 33 percent. The index of contract award prices of imported 256K DRAMs to OEMs fell, by the US International Trade Commission's figures, from 100 in September/October 1984 to as low as 7 in November 1985. The barb: US shipments of imported Japanese 256K DRAMs rose at the same time from 10.6 million units in 1984 to 28.0 million in January to September 1985, and their share of apparent US consumption of all DRAMs from 3.7 percent (January to September 1984) to 16.9 percent (January to September 1985) - while Mostek was shut down in October 1985 and Intel withdrew from the commodity 256K DRAM market.
What the sources leave open: OTA's Table C-1 labels its unit "parts per million"; with values around 1 that is doubtful, so the unit is reproduced here in quotation marks and not converted. Anderson's figures come from editor John Minck's foreword to Anderson's memoir and concern 1980, not 1985. No source describing incoming inspection of 256K DRAMs at a named plant in 1985 was found.