Intel’s new CEO, Pat Gelsinger, has wasted no time. While he won’t take over from Bob Swan until February, he’s already addressed Intel’s workforce in an all-hands meeting.
According to The Oregonian, Intel told assembled employees that it might wait to announce any change to its manufacturing plans until Gelsinger is onboard. This is an interesting wrinkle to the overarching manufacturing situation, and it implies that Intel wants a CEO with an engineering background to examine the situation and possibly make the decision. Then again, the Oregonian reports Intel “may” postpone a decision, not that it would, so the issue may still be under discussion.
Intel has pledged to decide the future of its manufacturing plans for next-generation chips before those CPUs would need to be in production by 2023. That’s the approximate date by which we’d expect 7nm CPUs to be in-market.
Currently, Intel is focused on Alder Lake and Rocket Lake. The former is Intel’s upcoming hybrid platform that will offer up to 16 cores with as many as eight full-sized CPU cores and eight low-power cores. There are rumors that we may see Intel hit some unusual thread counts here by supporting Hyper-Threading on the large cores but not the small ones.
Rocket Lake is Intel’s upcoming desktop platform that’s still built on 14nm but uses an updated microarchitecture based on last year’s Ice Lake mobile processors. Chipzilla has indicated Rocket Lake should deliver up to a 1.19x IPC improvement over and above Comet Lake.
Intel hasn’t officially revealed its 7nm product line, but the rumor mill believes it’s called Meteor Lake. At CES this week, Intel showed Alder Lake running in a laptop, implying it’ll be a mobile-first architecture. Alder Lake presumably comes to desktops in 2022, which clears the way for Meteor Lake in 2023. If Intel is building this chip in its own fabs, we can expect it to debut on 7nm. If it uses TSMC or licenses a TSMC process in its own factories, it might opt for 5nm or possibly 3nm depending on TSMC’s own technology ramp and node progress.
Intel can postpone that decision for a month or two, but it takes time to either implement another foundry’s process node or to design a chip specifically at TSMC. Whatever Intel is going to do, it has to start doing it soon.
Intel vs. ‘A Lifestyle Company’
Intel is taking the threat of Apple’s M1 very seriously. Gelsinger is said to have told employees, “We have to deliver better products to the PC ecosystem than any possible thing that a lifestyle company in Cupertino. We have to be that good, in the future.”
It’s good to see Intel taking the M1 seriously. Apple’s M1 chip hit the market like a bomb. While we’re happy to acknowledge that there are still a lot of questions about how Apple’s CPUs will compare with x86 across the width and breadth of the software market, the SoC is very good at what it does. Calling Apple a “lifestyle company” under these circumstances is a definite shot across the bow. Alder Lake should be the appropriate point of comparison for whatever higher-performance Mx CPU Apple debuts, so we’ll find out with that chip if Intel’s bravado is justified or not.
It’ll be very interesting to see what kind of mobile power consumption benefits Intel can extract from Alder Lake’s hybrid computing architecture, and we do expect Intel to continue improving x86 performance. Gelsinger’s attitude, however, is the right one. Both Intel and AMD need to respond to ARM’s encroachment with everything they’ve got, or risk x86’s long-term prominence in the desktop and laptop market. AMD has its own Ryzen 5000 mobile chips coming this year, with the expected 1.19x IPC uplift we’ve already seen on desktop chips. For now, Apple is the only company with an SoC that will realistically compete with either x86 company, but that could change in the future, depending on how the two manufacturers respond.
Microsoft CEO: Selling HoloLens to Military Is a ‘Principled Decision’
Microsoft CEO Satya Nadella dismissed accusations that the company is war profiteering by selling HoloLens to the US Army to increase troop lethality. It's not clear he's even considered the question.