Why Intel can’t quit x86

Here’s a nice perspective on Intel’s non-x86 efforts, and why they failed and x86 marched on, despite its weak points, and why Intel can’t quit x86.

Kudos for remembering that Intel made ARM chips.

Intel got into ARM by accident, when a court order forced Intel to buy DEC’s chipmaking business. One of the afterthoughts in that purchase was StrongARM, a cheap ARM device that ran at around 200 MHz and turned out to be a lot more popular than Intel ever imagined. Intel refined the design into its Xscale line, which was popular in its own right–many Windows CE PDAs used it, for example–but Intel gave up and sold the design to Marvell in 2006 as Windows CE’s popularity tanked.

The ironic thing is that ARM is a RISC architecture doing to Intel what x86 did to prior RISC efforts. Intel’s x86 chips weren’t better, and they usually weren’t faster, but they were cheaper. So one by one, vendors adopted Intel, either by switching to Windows, or porting their OS to Intel x86 CPUs. One system moved twice–Steve Jobs’ Nextstep moved from Motorola 68K to Intel, then to Motorola/IBM PowerPC after Apple bought it and turned it into OS X, then back to Intel again. If I had to wager, Apple had it running on Intel and AMD chips in labs the whole time, just in case. And now the rumor is they have OS X running on ARM in the labs behind the scenes, too. It wouldn’t surprise me.

Intel has momentum with x86–there’s plenty of stuff that runs on x86 on Windows, and nowhere else–but I think Intel made a mistake getting out of ARM. And I think AMD needs to get into ARM too.

The economics of selling $20 ARM SoCs is different from selling $300 Core i7s. The margins are lower, but the volume is potentially much higher.

I disagree with the thinly veiled assessment that Intel needs to retire x86–Windows on x86 is still a viable market. Even Itanium is still a viable market, albeit a niche one, but Intel is contractually obligated to keep making the chip at least through 2017. And even though it failed as the heir to x86, its sales are perfectly in line with specialty chips like IBM’s Power architecture. If Intel decides to divest itself of Itanium, I’m sure someone else wants to buy the rights and make it.

It’s possible that someday, ARM will overtake x86, and maybe someday even displace it. I won’t write the obituary just yet, though. Because 30 years ago, Motorola 68K was going to replace x86. Then 20 years ago, it was PowerPC and Alpha. And 10 years ago, it was Itanium. But the killer app for x86 never moved to the new chip, so x86 marched on, even as the killer app changed over the years.