A component test bench in a laboratory in Cambridge in Massachusetts, in front a grey foam tray with neat rows of dozens of identical small flat round cans of matt silver metal nine millimetres across, each with a gently domed plain smooth unmarked lid and a narrow base rim, from each can eight thin gold-yellow wires three and a half centimetres long hanging straight down or slightly spread, a hand in a white shirt sleeve lifting one can with tweezers toward a small bare steel box, behind them a test chamber of grey sheet steel with a small square glass window, cool fluorescent light from above.
What it was used for
Fairchild's Micrologic "G" element was a three-input NOR gate with 3 transistors and 4 resistors on one die. Eldon Hall at the MIT Instrumentation Laboratory made it the Apollo Guidance Computer's only logic type - one type instead of many, because reliability testing could then be pooled. After Robert Norman's visit in early 1962 Hall had large quantities bought to see whether the devices were any good; Fairchild filled an order for 100 within days. By early 1963 a prototype stood in 4 refrigerator-sized cabinets. For Block I the programme procured nearly 200,000 circuits, mostly from Fairchild.
What was new
What marked it out was the restriction to one type: with the same gate everywhere, test data could be pooled instead of spread over many packages. The price was quantity - about 4,000 circuits per computer and nearly 200,000 for Block I. The unit price fell from $43.50 in early 1962 to an average of $20 to $30 in Block I production, once epitaxy had improved yields. In 1963 the Apollo programme consumed 60 per cent of US integrated circuit production, by the Computer History Museum's account. Replacement came fast: from 1963 a denser dual-gate flat pack was developed for Block II, and the "G" element flew only in the Block I machines.
What the sources leave open: Both sources are Computer History Museum texts. For the figure per computer, the Silicon Engine page gives about 4,000 Type-G circuits per system, the CHM blog essay 4,100 IC packages without saying for which block. The figures 2,756 packages, 5,530 gates and 16,536 transistors belong there to Block II, that is to the later dual gate, not to the G element.
Connections
Every connection names its kind. “Forerunner” on its own would be a claim.