A leak from BenchLife last month confirmed that Intel (NASDAQ:INTC) planned to roll out a processor known as Kaby Lake in the 2016 time frame built on its 14-nanometer manufacturing technology in place of a 10-nanometer part known as Cannonlake.
It seemed reasonable to think that Cannonlake would have followed in the year after Kaby Lake. But a new report from KitGuru claims that Intel has outright cancelled Cannonlake and will, instead, follow Kaby Lake with Icelake, likely in the 2017 timeframe.
It could make sense
There are not too many details known about Kaby Lake, but there is evidence that it could incorporate some elements of the Cannonlake architecture (namely the graphics engine). That said, given that Cannonlake is/was a "tick" in Intel's "tick/tock" methodology, it is likely that the architectural improvements to the design would be fairly light aside from the new graphics/media engine.
Although moving from a 14-nanometer Kaby Lake to a 10-nanometer Cannonlake would likely bring some improvements in power efficiency due to the new manufacturing technology, Intel would be able to offer a bigger leap forward in 2017 by moving from a 14-nanometer part on an older architecture to a 10-nanometer part with a "new" architecture.
Is this plausible?
Intel has likely spent significant cash developing the Cannonlake processor core and the various chip projects that implement said core, which might lead some to believe that Intel will want to get a return on those designs before shipping Icelake.
But this, in my view, is flawed thinking.
The development costs associated with Cannonlake and the chips incorporating the core have likely been incurred over many years and are probably pretty much "sunk" at this point. Whether Intel actually sells Cannonlake parts come 2017 or Icelake parts won't have an impact on the company's revenue and profit.
In fact, if anything, Intel is better off bringing out the best designs that it possibly can in order to strengthen its competitive positioning and give system vendors more reasons to want to transition systems to the latest Intel chips as soon as possible.
This sounds better than what Intel did with Broadwell
Intel had well-publicized issues getting its 14-nanometer technology to acceptable manufacturing yields, which led to a delay in the company's Broadwell family of processors built on the technology. Ultimately, Intel chose to shorten the life of the Broadwell products and plans to bring out Broadwell's follow-on, known as Skylake, later this year.
If the report from KitGuru is true, then Intel will be able to do a more orderly transition to Icelake and 10-nanometer manufacturing technology. With the launch of 14-nanometer Kaby Lake in 2016, Intel will essentially get another year to improve its yields on its upcoming 10-nanometer manufacturing process. Then, once the new manufacturing technology is ready, Intel will be able to come out with a new architecture that really takes advantage of it.
In the meantime, Kaby Lake seems to bring some improvements that should make it a reasonable product for 2016. Additionally, since Kaby Lake will be a 14-nanometer part, and given that Intel's 14-nanometer technology should be quite mature come 2016, Intel won't have to worry about whether it can produce enough chips to support a broad product rollout.
Although this is just a rumor, as both an Intel investor and a PC enthusiast I hope this report is true. Intel's launch of Broadwell was not the company's finest moment, and if the company can avoid a similarly botched launch of Cannonlake, then all the better.