Arm Macs and AMD rising: How Intel’s endless 10nm struggles cost it so much

Intel’s endless 10nm nightmare has cost it so, so much.

It all started on September 5, 2014. That’s the day Intel introduced 5th-gen Core M chips based on “Broadwell,” the company’s first processors built using the 14-nanometer manufacturing process. Despite some manufacturing woes that pushed Broadwell back from its expected 2013 release, Intel’s offering served as the vanguard of processor technology. AMD remained stuck on the 28nm process with its abysmal Bulldozer architecture. A mere month later, the Apple iPad Air 2 launched with a custom A8X chip that couldn’t quite hang with Intel’s older Haswell CPUs in Geekbench—but it was getting close.

Nearly six years later, the tables have turned. Intel’s 10th-gen Core processors remain on an (upgraded) 14nm process. AMD’s Ryzen chips have snatched the computing crown, and Apple’s doing the unthinkable: switching Macs away from x86 CPUs onto its own custom Arm silicon. And if Apple’s flight from Nvidia GPUs after “Bumpgate” in 2009 is any indication, it won’t be coming back.

How did Intel get here? Let’s look at how the company lost its way, starting with the death of tick-tock.

The long road to 10nm

It wasn’t supposed to be like this. Intel’s original roadmaps expected 10nm chips to launch in 2016, with more advanced 7nm chips coming in 2018. Then the delays began.

The death of Intel’s vaunted “tick-tock” manufacturing process served as the canary in the coal mine. For years, Intel’s processors followed the tick-tock cadence, releasing upgraded CPUs with a smaller manufacturing process one generation, then a new microarchitecture built on the smaller process the following year. Tick-tock; tick-tock. The relentless innovation must have sounded like the doomsday clock to then-floundering AMD.

The troublesome 10nm process killed it. In early 2016, Intel confirmed that tick-tock was dead, adding a third leg to the process dubbed “optimization.” Intel’s 7th-gen “Kaby Lake” processors were flagged as the first “optimization” architecture in 2017, another 14nm chip following the releases of Broadwell and then Skylake. Considering that Intel still has yet to release 10nm desktop processors, it comes at no surprise that neither tick-tock nor tick-tock-optimization has been mentioned since.

Tick-tock’s demise obviously delayed 10nm’s arrival. Originally slated for a 2016 launch, by early that year Intel said that its first 10nm would be “Cannon Lake” in the second half of 2017, a die-shrink of the optimized Kaby Lake architecture. It wouldn’t launch until mid-2018, and only then in a handful of low-end systems with integrated graphics disabled. Later that year, we said “Cannon Lake is barely squeaked out in any reasonable volume.” The release went so poorly that when Intel previewed 10nm “Sunny Cove” cores to the press in December 2018, it also vowed to decouple its architecture and IP from manufacturing process as much as possible to prevent stalls like this from happening again.