At Intel's (INTC 1.77%) Technology and Manufacturing Day, the company showed information about its current-generation and upcoming chip-manufacturing technologies. Executives talked about the densities of these technologies as well as their performance and power characteristics.
What's interesting is that the company says that its upcoming 14-nanometer++ technology -- a third-generation version of the company's initial 14-nanometer technology that first went into production in the second quarter of 2014 -- delivers better performance than its upcoming 10-nanometer technology, as well as its upcoming 10-nanometer+ technology (though 10-nanometer+ seems to come extremely close to 14-nanometer++ in this regard).
Image source: Intel.
On the other side, Intel says that its 10-nanometer and 10-nanometer+ technologies can offer lower power consumption than its 14-nanometer technologies and, of course, much higher transistor density.
Considering how good Intel's 14-nanometer++ technology appears to be, is the company's first-generation 10-nanometer technology even worth using? Here's what Intel executive Kaizad Mistry had to say.
The answer? It depends
"Depending on the type of product, the lower power can be a significant advantage and a reason to go to 10 [nanometer] over 14+ or 14++," Mistry began.
He went on to say that products that are less power-constrained -- think high-performance notebook chips or desktop processors -- could benefit from using one manufacturing technology, while those more sensitive to power consumption might benefit from another.
Mistry also appeared to argue that it wouldn't just be low-power products that could benefit from potentially lower-performing (but lower-power and denser) 10-nanometer technologies, compared to 14-nanometer+ or 14-nanometer++.
"If you have an architectural innovation that significantly adds to the transistor count and adds to the power, that would provide performance and may benefit from being on 10 versus 14," he explained.
Translating this into Intel's product portfolio
In the second half of this year, Intel is expected to start selling the first products based on its 14-nanometer++ technology, known as Kaby Lake Refresh. These are aimed at high-performance laptops and are expected to have thermal design power starting at 15 watts and moving up from there.
Then, in the first half of 2018, Intel is expected to roll out higher-performance chips for desktops and larger or bulkier laptops, known as Coffee Lake, and also built on 14-nanometer++.
At the same time, Intel has said that it will begin shipping its first 10-nanometer chips based on its Cannon Lake architecture at either the very end of 2017 or the very beginning of 2018. These are likely to be targeted at low-power notebooks without fans.
The split between Kaby Lake Refresh and Coffee Lake, on the one hand, and the Cannon Lake product on the other seems like a good illustration of the idea that choice of manufacturing technology really does depend on the desired product characteristics.
That said, I do expect that Intel will begin migrating the entirety of its personal-computer processor portfolio over to some flavor of 10-nanometer -- likely 10-nanometer+ -- in the second half of 2018, with its Ice Lake architecture.
Even though 10-nanometer+ won't offer quite the performance that 14-nanometer++ will, per Intel's slides, the performance delta appears small enough that Intel should be able to deliver better products, through architectural innovations enabled by the power-reduction and density improvements of 10-nanometer+ over 14-nanometer++.
Foolish takeaway
It doesn't seem make much sense for Intel to transition the bulk of its products from 14-nanometer+ or 14-nanometer++ to the first-generation 10-nanometer technology. Performance on 10-nanometer is clearly lower than that on 14-nanometer++, and manufacturing yields on 14-nanometer++ are certainly going to be better than on 10-nanometer for a while yet.
However, once Intel begins production of its 10-nanometer+ technology -- which seems to offer nearly the performance of 14-nanometer++, while at the same time offering potentially lower power and much-improved transistor density -- it should be able to migrate the bulk of its product portfolio without too much issue.
