When Apple announced the new iPad Air last month, the most interesting thing about it wasn't its iPad Pro-inspired design or the bevy of new color options. No, it’s what the new Air had lurking inside that slim frame.
To our surprise, the 2020 iPad Air was the first device announced by Apple to use the new A14 Bionic chipset. That silicon’s impact won't be limited to tablets, either -- it will almost certainly power the next generation of iPhones, which Apple will unveil on October 13th. In a conversation with Engadget, Tim Millet, Apple's VP of platform architecture, and Tom Boger, senior director of Mac and iPad product marketing, shed some light on the company's approach to designing the A14, and what it means for the iPad Air and beyond.
Building a beast
At a high level, the A14 seems similar to Apple’s other Bionic chipsets. This system-on-a-chip packs a six-core CPU — two cores high-performance cores and four for lower-priority tasks — just as the A12 and A13 did. The number of GPU cores here has also remained unchanged at four. Don't be fooled by these passing similarities, though: Because the A14 was designed for a 5nm manufacturing process, there’s more going on in this system-on-a-chip than ever before. But let’s take a step back first. The shift to ever-denser chipset designs has been happening for years, and shows no sign of slowing.
The A14 might be the world’s first commercially available 5nm chip, but Apple’s rivals aren't far behind. Qualcomm first 5nm mobile chipset, the Snapdragon 875, could debut as soon as December at the company’s virtual Snapdragon Summit. And then there’s Samsung, which — in addition to manufacturing those Snapdragons for Qualcomm — has begun pulling back the curtain on its 5nm Exynos 1080 chipset.
The main benefit of chips based on these new manufacturing processes is that they’re more densely packed with transistors, incredibly small switches that can control the flow of electrons. These serve as the foundation for logic gates, which beget integrated circuits, which beget full-blown processors.
In any case, the shift to 5nm meant Apple had far more transistors to devote to all the systems on the chip. Think: 11.8 billion, up from the 8.5 billion the company had to work with in last year’s A13 Bionic. As you’d expect, that huge uptick in transistor count gave Apple the extra processing bits needed to build significantly faster, more efficient CPU and GPU cores. But it also gave Apple the latitude to make more subtle improvements to a device’s overall experience.
“One of the ways chip architects think about features is not necessarily directly mapping [transistors] to a user feature in the product so much as enabling the underlying technology, like software in the graphics stack to be able to leverage a new capability in the GPU,” Millet said. “That will inevitably come as a visual feature in a game, or in a snappy transition in the user interface.”
The switch to a 5nm design for the A14 also gave Apple the latitude to devote more of its transistor “budget” to components beyond just the CPU and GPU. And when it comes to achieving the best all-around experience, companies like Apple, Samsung and Huawei -- the only other companies to design chips for their own mobile devices -- have a distinct advantage. In this case, because Apple exercises full control over what goes into its systems-on-chips, it can invest in additional processing cores and components before they go mainstream.
The best example is the company's Neural Engine, a component that debuted in the iPhone X's A11 chipset to accelerate the sorts of neural networks needed for features like secure face unlocking, voice recognition for Siri and augmented reality, among other things. Apple was among the first to integrate a dedicated neural accelerator into its chips — Huawei announced the Kirin 970 and its neural processing unit a week before Apple revealed its own Neural Engine, and Samsung and Qualcomm only caught up later.
Unsurprisingly, this year's Neural Engine is a far cry from the first one we saw in 2017. While that original co-processor could perform 600 billion operations per second, last year's A13 raised the bar to 6 trillion operations in the same amount of time. Meanwhile, the A14 generally obliterates the bar by performing a claimed 11 trillion operations per second.
That boost was made possible by a big redesign: The A14's Neural Engine now packs 16 cores, compared to eight in last year's A13. Doubling the engine's core count was an interesting choice since many of the iOS features that relied on it already seemed to run well enough. Since that’s the case, why not instead devote more of those new transistors to further ramping up CPU and GPU performance, which most people may more immediately notice?
The answer is two-fold. For one, Apple continues to see huge potential in supercharging neural networks, not just for the sake of its own software experiences, but for the ones app developers might be able to achieve with the right components in place. The popular image editing app Pixelmator Pro, for instance, leans on the Neural Engine for a feature that makes low-resolution images look surprisingly crisp and clean. Meanwhile, on the other end of the creative spectrum, Algoriddim’s djay Pro AI app uses the Neural Engine to more capably isolate vocals and instrument tracks in songs.
“We saw the opportunity to do things that would have been impossible to do with a conventional CPU instruction set,” Millet said. “You could in theory do many of the things the Neural Engine does on a GPU, but you can't do it inside of a tight, thermally constrained enclosure.”
And that’s a nice segue to the other half of the answer, which is that Apple had to balance sheer horsepower with efficiency. After all, there's no point in making sure the horses run fast if they tire out too soon.
“We try to focus on energy efficiency, because that applies to every product that we build,” said Millet. By making that a fundamental focus of its chip designs, Apple doesn’t have to worry about a situation in which it “focused on energy efficiency for the phone [in a way] that’s not going to work in an iPad Air. Of course it’s going to work.”