[Delk]

I'm guessing that was a 286. I think Intel parts topped out at 12.5 MHz but AMD and Harris eventually reached 20 or even 25 MHz. I still have my original PC with a 12.5 MHz one.

The difference with the 386, I think, is that AFAIK the second-sourced 8086 and 286 CPUs from non-Intel manufacturers still made use of licensed Intel designs. The 386 (and later) had to be reverse engineered again and AMD designed their own implementation. That also meant AMD was a bit late to the game (the Am386 came out in 1991 while the 80386 had already been released in 1985) but, on the other hand, they were able to achieve better performance.

[rasz]

AMD didnt clean room 386, nor even 486. AMD directly copied Intel microcode 100% 1:1 for 386, and later admitted to copying parts for 486 (smm? ice?). Sept. 4, 1993 LA Times article:

>AMD said Friday that its “independently derived” 486 microprocessor borrowed some microcode from Intel’s earlier 386 chip.

Borrowed hehe. Ended up in a 1995 settlement where AMD fully admitted copying and agreed to pay $58mil penalty in exchange for official license to 386 & 486 microcodes and infamous patent 338(mmu). Intel really wanted a legal win confirming validity of their patent 338 to threaten other competitors. 338 is what prevented sale of UMC Green 486 in USA. Cyrix bypassed the issue by manufacturing at SGS and TI who had full Intel license https://law.justia.com/cases/federal/district-courts/FSupp/8...

>were able to achieve better performance

Every single Am386 instruction executes at same cycle count as Intel counterpart, difference is only official ability to work at 40MHz.

[dspillett]

> I'm guessing that was a 286.

It is, yes. I meant to mention that detail!

> The 386 (and later) had to be reverse engineered … That also meant AMD was a bit late to the game

There were also legal matters that delayed the release of their chips. Intel tried to claim breach of copyright with the 80386 name¹ and so forth, to try stymie the competition.

> they were able to achieve better performance.

A lot of that came from clocking them faster. I had an SX running at 40Hz. IIRC they were lower power for the same clock then Intel parts, able to run at 3.3V, which made them popular in laptops of the time. That, and they were cheaper! Intel came out with a 3.3V model that had better support for cache to compete with this.

--------

[1] This failed, which is part of why the i386 (and later i486 and number-free names like Pentium) branding started (though only in part - starting to market direct to consumers rather than just EOMs was a significant factor in that too).