Most launch events for new technology products have a section to discuss the digital innards. This is strictly for the geeks: a chance to wallow in what are known as the “speeds and feeds” of the technology. It used to be the centrepiece of any new product announcement, until Steve Jobs shifted the industry’s attention to user experiences.
But to understand how competition is shaping up in the smartphone world, it was worth paying attention to the digital guts during this week’s iPhone launch.
Apple’s latest handsets are the first devices anywhere to include a chip made with 7 nanometre process technology, meaning that the width of the features etched on to the silicon has reached a new level of miniaturisation, down from the previous 10nm. The chips are designed by Apple but manufactured by TSMC.
First, it is worth noting what a feat this represents in chip process technology. Getting to 7nm has been much tougher for the industry than previous shrinkages, and a sign of how Moore’s Law — which predicted regular advances in the number of transistors that can be squeezed on to a chip — is running out of steam.
Last month GlobalFoundries, the chip manufacturer controlled by Abu Dhabi, put off its own plans for 7nm indefinitely. Intel, whose long leadership of the chip industry was built on its manufacturing prowess, has also been struggling; after several delays, products containing comparable chips from the US chipmaker will not be available until late next year.
Like all “shrinkages”, as the feature size on a chip takes the next step down, 7nm technology translates directly into more processing oomph and longer battery life. The new XS Max — the most powerful and expensive iPhone Apple has ever made, with the biggest screen — promises a full hour and a half of extra use from each charge.
But there is much more to this than size. Apple’s chip includes a specialised accelerator for machine learning known as a neural processing unit.
With the promise of applications that can learn from masses of data, this is where much of the effort in new hardware design is now focused. Developers are promised a chip that run their apps nine times faster, but using only a tenth of the power, as the NPU in last year’s iPhone.
This is the culmination of work that Apple has been doing for years. What is most surprising, according to chip analyst Patrick Moorhead, is that the company has clawed its way to the forefront of the chip industry “without any apparent, major execution issues”. He says Apple is now “world-class” in developing the blueprint for this sort of system on a chip.
This points to a race that has broken out among the top smartphone makers. Huawei, which also relies on TSMC to produce to its own designs, stole a march on Apple by unveiling a 7nm chip two weeks ago — though the first handsets will not ship until a month after Apple hits the market. Like Apple, Huawei unveiled its first NPU last year. China Inc may be far behind the US in terms of its semiconductor capabilities, but this is one important area that looks to be getting closer to parity. Samsung is also in the NPU race.
For the handset makers, taking machine learning out of the cloud and putting it into users’ hands has clear advantages. It means less need to shuttle large amounts of data back and forth to cloud data centres. That puts less strain on bandwidth and, more importantly, brings instant response time for applications. It also means built-in privacy because a user’s data never needs to leave the device.
Importantly, Apple’s new chip — called the A12 Bionic — will be included in the new, lower-priced iPhone XR. That should put it in the hands of many more users, in turn giving developers greater incentive to create applications that take advantage of the technology.
This is the biggest question left after this week’s iPhone launch. The technology has leapt ahead: how will it change what people can do with their phones, and hence how competition between phonemakers will develop?
Apple’s Face ID was one of the first applications of its NPU last year. This facial recognition technology will now be much more widely available to unlock handsets and act as an authentication for Apple Pay — something that could presage a big push by Apple to get wider adoption for its payment system, according to Goldman Sachs.
Most importantly, though, Apple will be counting on other app developers to dream up compelling new uses for the technology. A powerful new AI machine is due to start shipping next week: the next phase of the smartphone wars has begun.