Why so long a lag between intro of 286 & 386 and first PC's?


Feb 18, 2016
I was looking at info on the early PC computers and processors, and I was struck by how the 286 was introduced in February 1982, but the IBM PC/AT did not come out until August 1984, and the 386 was introduced in October 1985, but the first PC to use it, the Compaq Deskpro 386, was not announced until September 1986. I'm wondering why the long lags between the chips becoming available and the computers that used them. It's not like Intel had a lot of other markets to sell those chips to besides the IBM-compatible PC market, so the factories were just sitting idle? Obviously in today's market the new-gen chips make it to the market much faster, but I'm wondering if back then the performance gains were so much greater, that the amount of engineering for the supporting motherboard and chipsets was more substantial than today's, where the same sockets and other motherboard standards are carried over from one gen CPU to the next.
There were several things that caused delays in adoption of the 386.

- Early 386 parts had poor yields, forcing Intel to introduce them at 12.5 MHz, instead of the original 16 MHz. At these speeds, there was no benefit of the 386 over the 286 because virtually everything was 16 bit x86 code, and the 386 was no faster than a 286 at 16 bit code.
- Early 386 mask revisions had a bug that caused incorrect results when using 32 bit code, leading Intel to have to re-bin their entire inventory. Defective chips were marked with "16 bit S/W ONLY" and working chips were marked with a double sigma.
- The 386 bus was tri state multiplexed, which increased costs and limited the performance of the CPU.

Since there was little benefit to a 386 over a 286, the market just ignored it until market conditions were more favorable. It made no financial sense to buy an extremely expensive CPU that no software runs on natively, and doesn't offer any speed benefit over your existing 286.