22 years ago, on September 23, 2003, AMD changed the game for x86 once and for all. They released the Athlon 64 CPU, a chip that did something Intel didn’t want. Intel didn’t want to extend x86 to 64 bits. But when AMD did it, it forced Intel to clone AMD, rather than the other way around.
Why Intel didn’t want to go 64-bit
Even in 2001, x86 had decades of baggage attached to it. It was a 32-bit architecture that had been extended from a 16-bit architecture. But that in turn had been extended from an 8-bit CPU design from 1972 that, believe it or not, originated at Datapoint, not Intel.
This was great for backward compatibility. 8-bit applications were very easy to port to x86 in the early 1980s, and those early DOS applications still ran flawlessly on modern systems 30 years later. For that matter, it’s not impossible to get them running even today.
Removal of the ability to run 16-bit applications in 64-bit Windows was a design decision, not a technical limitation.
Intel wanted to start over to go 64-bit. Without having to worry about backward compatibility, they could design something that would be faster and more efficient. In theory at least, it would be able to scale higher in clock speed. And there was no question a new design would outperform a theoretical 64-bit x86 when running at the same speed because of efficiency.
And if you are cynical, there was one more motivation. If Intel could start over, they wouldn’t have to worry about competing CPU designs, at least not for a very long time. The new design would be encumbered with so many patents, it might be 20 years before someone could clone it.
Keep in mind that in 2003, not only was AMD in the picture, but Transmeta was still in the picture, and Cyrix was fading but not completely gone.
Starting over with a new CPU architecture outright was massively attractive to Intel.
This new 64-bit architecture wasn’t theoretical, either. Intel was producing it. It was called Itanium, and Intel first released it in June 2001.
... continue reading