Intel's P6 derivatives have been around for a while - can Keller revitalize x86 for the next decade?
Jim Keller faces what may be the most interesting challenge of his storied career as a microprocessor engineer-reshaping the Intel x86 for the 2020's.
In part through the luck of being tied Microsoft’s operating systems, Intel’s processor architecture came to dominate the PC. Through years of its own hard work, Intel unseated a host of powerful, entrenched Linux server processors to dominate the data center, too.
But with the exception of a solid foothold in iPhone modems today, the x86 missed the smartphone boom. Now it finds itself at the start of an era where machine learning is rewriting the rules of computing in ways that could reformat the landscape of microprocessors.
Past leaders are routinely unseated in technology transitions as new leaders emerge. Fighting that tide will take all the skill and luck Intel’s managers and engineers can muster.
So far, chief executive Brian Krzanich has done a decent job of turning the giant corporate cruise ship to face the coming waves of change. He brought in Venkata "Murthy" Renduchintala from Qualcomm in late 2015 to shake up the Intel organization. More recently, he acquired Movidius, Nervana, MobilEye and Altera, giving Intel an arsenal of AI architectures.
Now someone needs to make sense of what goes where, when and how. That’s a great job description for Keller, who left Tesla to join Intel in late April.
Keller led the Zen x86 core design that is now bringing AMD back to profitability, as well as its K8 Athlon years earlier. In between, he worked at two chip startups, did a tour of duty at Broadcom and designed SoCs for iPhones and iPads while at Apple.
But as they say, past performance is no guarantee of future success. Intel’s cultural momentum is a huge force to fight. It’s still not clear how AI will reshape computing. And other big forces are at play.
Intel also faces the slowdown in CMOS scaling. The main engine of past growth for Intel and the whole semiconductor industry is now showing its age as even Intel’s own fabs struggle to ramp a 10nm node.
Meanwhile, the chip market is headed into an era of hyper-fragmentation. No one system has emerged to approach the size and significance of the PC and smartphone. So, tomorrow’s companies will have to struggle for design wins in many different sectors simultaneously — from IoT end nodes and smartwatches to self-driving cars, data center networks and everything in between.
It’s quite a set of interwoven challenges that I have no doubt folks like Jim Keller understand well. Like many others, I have the privilege of sitting in the stands, fascinated to watch what will emerge in the next three to five years before Keller finally retires — or moves on to his next big challenge.
— Rick Merritt, Silicon Valley Bureau Chief, EE Times