Re: Nice about Andy
"The memory business was simply lost as the Japanese production technique become superior."
I have a late 1970's electronics magazine that contains an article titled "A little less witchery and a little more craft"(*) that gives the stat that in 1978 wafer yields for even bog standard TTL stuff was around 2% and for CPUs it was sub-1%
Yes, that does mean 98 out of every 100 devices on the wafer were duds (each wafer was only about 4 inches back then - check the BBC "This changes everything" Horizon documentary from 1977 if you can find it.)
Part of the reason for the proliferation of transistor types was down to the way they were made - each batch was characterised after manufacture and if they fell outside of the posted tolerances for any given type (which were generally just defined by the variations within any given batch) a new part number and spec would duly be conjoured up to match the batch's characteristics.
Better manufacturing techniques (pioneered by the japanese makers) resulted in extremely tight consistency between batches for individual transistors and increasingly rigid cleanliness standards resulted in much higher product yields for wafers.
Intel was pretty late to the party when it came to getting its yields up but these days their production facilities are amongst the best in the world.
(*) The article was in Electronics Australia, but it was a reprint from a US magazine whose name escapes me.