For most of the past decade, Intel has been releasing new chips in a tick-tock pattern. A tick means the shrinking of the manufacturing process technology, while a tock signals an improved microarchitecture.
But for the first time since 2007, Intel is delaying the tick. The company’s next two families of desktop and notebook chips will be based on the same 14nm process as today’s Broadwell processors.
Intel doesn’t expect to move to 10nm chips until the second half of 2017.
The company had already planned to release next-gen chips, code-named Skylake, starting later this year. During an earnings call, the company acknowledged that process shrinks are taking a bit longer than they used to: instead of moving to a smaller process every two years, Intel is on track to do it every 2.5 years.
So the company says it’ll introduce a 3rd chip family based on 14nm designs in 2016.
It’s code-named Kaby Lake, and that’s about all Intel is saying so far. But news of the new chips leaked a few weeks ago, suggesting that Kaby Lake chips will range from 4.5 watt Y-series processors (which will possibly be branded “Core M,”) to 80 watts desktop chips.
Does this mean Intel’s 2016 chips won’t be much better than the company’s 2015 models? It’s a little too early to say, but Kaby Lake chips be based on the same architecture as the Skylake chips which will launch later this year. As AnandTech notes, Intel will plan to offer some improvements that could lead to better performance or new features… which may help PC makers convince you not to wait another year before buying a new computer.
But odds are that if you’ve got a Skylake-powered laptop, you probably won’t feel the need to buy a Kaby Lake model to replace it.
There have been a lot of headlines suggesting that this change means the end of Moore’s Law, but that’s always been more of an observation than a law… and it’s been changed before.
Intel co-founder Gordon Moore noted in 1965 that there seemed to be a doubling of the number of components packed into integrated circuits every year and predicted that this would continue… until 1975 when he adjusted his outlook to every two years. Others have modified the time to 18 months, saying that chips double in actual performance every year and a half thanks to other improvements in addition to the number of transistors.
So it’s not surprising that we’re reaching a point where chip makers are having a hard time doubling performance, transistor density, or anything else every 12, 18, or 24 months. The fact that Intel still thinks it can do it every 30 months or so is still intriguing — but then again, IBM just demonstrated that it’s possible to build a 7nm chip, so even if it’s taking Intel longer than planned to hit 10nm, it doesn’t look like we’re going to hit the end of the road anytime soon.
via Ars Technica