Apple will launch the newest iPhone 12 series today. It is expected that the Apple iPhone 12 series will be powered by Apple's A14 Bionic chip which the company first announced with the iPad Air 2020 update last month. The A14 Bionic chip packs a six-core CPU, with two high-performance cores and four cores for lower-performance tasks, alongside a quad-core GPU. The A14 Bionic is designed on a 5nm manufacturing process, making it the first commercially available 5nm chip. Ahead of the iPhone 12 launch, Apple's VP for platform architecture Tim Millet has shared more insight on designing the A14 Bionic chip and the impressive performance they have seen from the A14 Bionic. In an interview with Engadget, Millet, along with Apple Senior Director for Mac and iPad Product Market, Tom Boger shed some light on Apple's plans for the A14 Bionic chipset, and what it means for the company.
The transition to a 5nm chip meant Apple now has more transistors to devote to all the systems on the chip. The A14 Bionic has 11.8 billion transistors, up from the A13's 8.5 billion transistors. While the extra transistors directly give Apple more space to put faster CPU and GPU cores, it also gave Apple the base to make more improvements to a device's overall experience. Further, the 5nm architecture gave Apple the base to devote more transistors to things beyond the CPU and GPU. This gives Apple the option to decide whatever goes into its SoCs.
This is where Apple choses to put the Neural Engine. While the first Neural Engine we saw on the iPhone X in 2017 could perform 600 billion operations per second, the Neural Engine on the A14 is capable of performing a claimed 11 trillion operations per second. Apple made this possible by crunching in 16 cores within the A14's Neural Engine, as compared to eight in last year's A13 Bionic chip. Now, while doubling the number of cores on the Neural Engine is an impressive feat in itself, many iOS features that relied on it seem to run seamlessly without the latest A14 Chip's Neural Engine. So why not simply use the extra space for ramping up CPU and GPU performance? Millet told Engaged the answer to this question is two fold.
One, Apple continues to see huge potential in supercharging neural networks, not just for the sake of its own software, but for what developers might be able to achieve with the right components in place. The popular image editing app Pixelmator Pro, for example, uses the Neural Engine for a feature that makes low-resolution images look more crisp. “We saw the opportunity to do things that would have been impossible to do with a conventional CPU instruction set. You could, in theory, do many of the things a Neural Engine does on a GPU, but you can't do it inside of a tight, thermally constrained enclosure of a conventional GPU.” Millet said.
Further, Millet said that the other part of the reason for using Neural Engines is that Apple had to balance the power with efficiency. “We try to focus on energy efficiency, because that applies to every product that we build. 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 iPhone in a way that’s not going to work in an iPad Air. Of course it’s going to work," Millet told Engadget.
While all of this seems super impressive from the face of it, it is still a matter of speculation as to how fast exactly is the Apple's A14 Bionic chipset from its predecessors. Apple has not given any details on the performance figure as compared to the A13 Bionic. During the iPad Air presentation as well, Apple had said that the A14 Bionic allows a 40 percent faster performance and 30 percent better graphics than the previous model. That comparison, however, was for the previous iPad Air's A12 Chipset. Even then, the real-life performance figures vary from what Apple claims. Boger told Engadget that those figures are an amalgamation of "real-world application workloads." To put it more simply, they're the numbers derived from different performance factors to demonstrate what it’s like to actually use the A14 Bionic chip.
Millet, on the other hand said that Apple understands that single-thread performance for a lot of applications is important. Hence, the company makes sure that when it's talking about things like that, it is representing the single-thread performance well.
When asked about the A14 Bionic blurring the lines between the iPhone and the iPad, both Miller and Boger pointed out that the current A12Z chipset on the iPad Pro has more CPU and GPU cores than the A14 Bionic. The bigger GPU computing power basically means that the iPad Pro will continue to be better suited for graphical work and other high-performance tasks. Boger, however, noted that the A14 does actually outperform the A12Z in a few areas.
Millet also said that when Apple starts working on a chipset, they take into account the company's entire lineup. "Ultimately, we want to make sure that when we build a CPU for one generation, we’re not building it necessarily only for one," he said. When asked if the development for Apple's A14 chips was influenced by the company's work on Mac chips, Millet said, "sometimes, it's the constraints of a unique platform that drive intention."