Is this the death of Moore's Law? Intel delays Cannonlake to 2017

"The lithography is continuing to get more difficult," admits CEO Brian Krzanich
Author:
Publish date:
Social count:
0
1-intel-cannonlake-delay-moores-law.jpg

Intel has delayed its 10-nanometer (10nm) Cannonlake chip technology from 2016 to 2017, effectively breaking its own 'Moore's Law'.

Moore's Law is an observation from Intel co-founder Gordon Moore that states the number of transistors on a chip roughly doubles every two years.

Moore first stated this back in 1965, so it's ironic that as Intel celebrates Moore's Law's 50th anniversary in 2015, it could also be the same year it is broken.

Back in 2013, Intel said it could get processors fabricated using a 10nm process by 2015. It later delayed this Cannonlake tech to 2016, and has now pushed it back to the second half of 2017.

However, with Intel's 14nm Skylake processors due later this year, it means there's a gap to fill in 2016, where Cannonlake would have launched. To fill this gap, Intel will release a refined version of Skylake, another 14nm series, codenamed Kaby Lake.

This means, with Broadwell from 2014, Skylake this year and Kaby Lake in 2016, Intel will have launched three families of 14nm chips in three years.

This breaks Intel's 'tick-tock' cycle of reaching a new process size one year, then launching a new micro-architecture based on that size a year or so later, and so on.

Intel CEO Brian Krzanich admitted that Moore's Law may be changing. He said that getting to 10nm is difficult, and this in turn is preventing Intel from working on 7nm sooner.

He said, during an earnings call, as transcribed by Seeking Alpha: "Just last quarter we celebrated the 50th anniversary of Moore's Law. In 1965, when Gordon's paper was first published, he predicted a doubling of transistor density every year for at least the next 10 years. His prediction proved to be right. And in fact, in 1975, looking ahead to the next 10 years, he updated his estimate to a doubling every 24 months.

"These transitions are a natural part of the history of Moore's Law and are a by-product of the technical challenges of shrinking transistors while ensuring they can be manufactured in high volume.

"As node transitions lengthened, we adapted our approach to the tick-tock method, which gave us a second product on each node. This strategy created better products for our customers and a competitive advantage for Intel. It also disproved the death of Moore's Law predictions many times over. The last two technology transitions have signaled that our cadence today is closer to 2.5 years than two.

"To address this cadence, in the second half of 2016 we plan to introduce a third 14-nanometer product, codenamed Kaby Lake, built on the foundations of the Skylake micro-architecture but with key performance enhancements.

"Then in the second half of 2017, we expect to launch our first 10-nanometer product, code named Cannonlake. We expect that this addition to the roadmap will deliver new features and improved performance and pave the way for a smooth transition to 10-nanometers."

When pressed specifically on the delay of the 10nm tech, Krzanich added: "It's similar to what happened on 14-nanometer. Remember, on all of these technologies, each one has its own recipe of complexity and difficulty, 14-nanometer to 10-nanometer same thing that happened from 22-nanometer to 14-nanometer.

"The lithography is continuing to get more difficult as you try and scale and the number of multi-pattern steps you have to do is increasing. This is the longest period of time without a lithography node change. 

"And then you look at the pattern we've been having with the same kind of sets of conditions, which was the 22-nanometer technology and the 14-nanometer technology. And we said those took about 2.5 years. When we go from 10-nanometer to 7-nanometer, it will be another set of parameters that we'll reevaluate this.

"We'll always strive to get back to two years. We believe our lead in Moore's Law will not change dramatically."

However, in 1965, when Gordon's paper was first published, he predicted a doubling of transistor density every year for at least the next 10 years. But in 1975, looking ahead to the next 10 years, he updated his estimate to a doubling every 24 months, as Krzanich pointed out.

So in Intel's defense, maybe Moore's Law isn't dying - it's just changing to move with the times.

Back in 2013, it was predicted that Moore's Law could end as early as 2020, because of economics, not physics.

Related