An Exclusive Look Inside Apple’s A13 Bionic Chip

Here’s how the chip in each new iPhone works and what it tells us about the future of mobile technology.
iPhone 11 Pro
Photograph: Apple

Application

Hardware

Company

Apple

End User

Consumer

Technology

Chips

About 72 minutes into the annual iPhone launch event, Apple senior vice president of marketing Phil Schiller invited Sri Santhanam to come onstage and talk about the brand-new A13 Bionic chip found inside all three of the new phones. The slight and shy Santhanam, Apple's vice president of silicon engineering, then spoke for four minutes. In many ways, they were the four most important minutes of the entire event. Not that anyone noticed—the audience was seduced by the shiny new iPhones, the three-camera system, the magical Night Mode, the impressive video capabilities and, more importantly, the boost in battery power.

By the time Santhanam was done talking, all I could think of were the numbers. Apple’s new chip contains 8.5 billion transistors. Also, there are six CPU cores: Two high-performance cores running at 2.66 GHz (called Lightning), and four efficiency cores (called Thunder). It has a quad-core graphics processor, an LTE modem, an Apple-designed image processor, and an octa-core neural engine for machine intelligence functions that can run over five trillion operations per second.

This new chip is smarter, faster, and beefier, and yet it somehow manages to consume less power than its predecessor. It’s about 30 percent more efficient than last year’s A12 chip, one of the factors that contributes to the extra five hours per day of battery life in the new iPhones.

The launch of the iPhone 11 Pro and its siblings only reaffirms that Apple's real advantage over its competitors comes from owning the entire vertical stack: the software, the system hardware, and the chip design. You can see the benefits of these gains in the iPhone’s feature set, from its augmented reality capabilities to its computational photography modes like Deep Fusion and Night Mode.

"One of the biggest examples of the benefits of the performance increase this year is the text to speech," Schiller said when we sat down to talk about A13 Bionic and its capabilities. "We've enhanced our iOS 13 text-to-speech capabilities such that there is much more natural language processing, and that's all done with machine learning and the neural engine."

Clock Cycles

Apple has come a long way from the launch of the original iPhone in 2007. That first handset was slow and unable to perform even the most basic tasks like copying and pasting text. It had terrible battery life. Its camera would make a supermodel look like the Bride of Frankenstein. Multitasking was almost nonexistent in the original iPhone, which was powered by a chip that ran at 412 MHz. The handset was pieced together from components that included a chip used in Samsung DVD players. It was hard to imagine that such a device could one day upend the entire idea of phones, computing, and communication.

It quickly became apparent to Apple that it would need to build the entire stack—soup to nuts—if it wanted to stay ahead of its competitors, especially those in the Android ecosystem. Apple’s decision to design and build its own silicon was made sometime in 2008. At the time, the company had a mere 40 engineers working on integrating chips from an assortment of vendors. Then, in April of 2008, Apple bought a chip startup called P.A. Semi for $287 million. That increased the total number of chip engineers to about 150 and brought home expertise on what matters most on a phone: power efficiency. The fruits of this group’s labor were first revealed to the world in the iPad 4 and the iPhone 4. Those devices were powered by a processor named A4, which was a modified version of a chip design from ARM Holdings. The A4’s primary focus was to make the Retina displays shine.

Over the years, the Apple chips have enabled features that cause the majority of oohs and aahs at its famous events. Siri, video calling, fingerprint- and image-based identification, the camera’s many powers—all the result of the silicon progress made by Apple. At the 2017 launch of the iPhone X, I wrote on my blog, "FaceID is a perfect illustration of Apple's not-so-secret ‘secret sauce’—a perfect symbiosis of silicon, physical hardware, software, and designing for delight. Their ability to turn complex technologies into a magical moment is predicated on this harmonious marriage of needs." This is Steve Jobs' real legacy for the company he cofounded.

The Heat Is On

Johny Srouji runs Apple's sprawling chip operation along with other hardware technologies. Many believe that a big chunk of the company’s annual research-and-development budget is earmarked for Srouji’s team. "Steve came to the conclusion that the only way for Apple to really differentiate and deliver something truly unique and truly great, you have to own your own silicon," Srouji told Bloomberg Businessweek a few years ago. The company is said to have a few hundred members in its chip operation, but press Apple executives for specifics, and they clam up fast.

Apple's chip advantage didn't go unnoticed in the industry. Using merchant silicon wasn't enough to catch up with Apple, which kept hammering its chip advantage, one phone and one tablet at a time. Huawei and Samsung—the latter being a frenemy of Apple from the very beginning—are two companies that quickly realized that the future of mobile technology was going to require custom silicon that allowed them to stay ahead of their Android rivals and better compete with Apple.

Apple VP Sri Santhanam talks about the A13 Bionic chip onstage last week at the Steve Jobs Theater in Cupertino, California.

Photograph: Apple

These companies, along with Qualcomm, are in a silicon arms race, constantly shuffling slots on the leaderboard. The last-generation A12 Bionic chip owned a slight edge over Apple’s rivals when it was announced, and then this year, Apple took the occasion of its iPhone 11 launch event to reinforce its lead.

Linley Gwennap, the founder of the research consultancy The Linley Group and publisher of the influential Microprocessor Report newsletter, is widely regarded as one of the foremost processor experts. Gwennap has spent most of his life dedicated to processors and chips, and is not as easily impressed by marketing speak. Sure, Apple has an advantage, he says, and it wins on benchmarks. But the edge isn't that much.

When talking about the previous-generation A12 Bionic in an interview, Gwennap points out that while Apple leads the single-CPU race, the others are relatively competitive with them.

"I don't see them as far ahead," he says. "I would expect Samsung, Qualcomm, and Huawei will up their game."

So have they stepped up their game since last-year’s A12? Exactly how does the new six-core A13 Bionic stack up against the latest chips from Apple’s three big rivals? Let’s look at the numbers.

Samsung's newest processor, the Exynos 9825, has eight cores arranged in three clusters: two high-performance custom Mongoose cores running at 2.73 GHz, another two Cortex A75 cores running at 2.4 GHZ, and four efficiency-focused Cortex A55 cores running at 1.9 GHz. There is a Mali GPU and Samsung's neural processing unit, along with LTE and memory capabilities.

Huawei's chip, called the Kirin 990 5G, follows a similar tri-cluster, eight-core (also known as octa-core) approach. There are two high performance Cortex A76 cores running at 2.86 GHz, another two A76 two-cores running at 2.35 GHz, and four efficiency-focused Cortex A55 cores running at an even slower 1.95 GHz. Rounding out the chip is a 16-core GPU and a Da Vinci neural engine with three cores. Huawei's chip contains a whopping 10.3 billion transistors.

Qualcomm's new Snapdragon 855 Plus is very much like the Kirin 990 and Exynos. It uses custom Kryo 485 Gold cores with one powerful cluster clocked at 2.96 GHz, another three Kyro 485 Gold cores running at a clock speed of 2.42 GHz, and four efficiency-focused Kryo 485 Silver cores running at 1.78 GHz. It includes an Adreno GPU and Qualcomm's Hexagon 690 AI engine.

Those chips have some faster components and more of them, so you may think those chips perform better than Apple’s. But the reality is that we hardly use the entire capacity of the chips that come in our mobile devices. One or two high-performance cores are enough for most of what we throw at our phones. Apple's six-core design might seem lagging compared to the eight-core processors from the competitors, but really, the two big processors on its chip easily outperform its rivals’ designs. Apple’s processors consume power more efficiently, and that gives them a distinct advantage over competitors. For instance, Samsung's Mongoose chips need to be used judiciously, lest they cause the device containing them to overheat. Even the newly designed custom efficiency cores in A13 also best their competitors.

"Although Apple's cores aren't the biggest, they continue to lead in mobile performance," noted Gwennap earlier this year in The Microprocessor Report. And at the time he wrote that, he was talking about the A12 chip. The A13 performs about 20 percent better.

So the takeaway here is that specs and benchmarks don't take into account Apple's real advantages—tight integration into the device and the company's development strategy for squeezing more runtime out of its batteries while boosting the performance of key apps.

Power Play

So, how does a phone company illustrate these technical gains in a way that resonates with customers? The chip-speak doesn't matter. What matters is having the best camera, the fastest phone, and—oh yes—the biggest battery. The longer we get to use Instagram, Facebook, or YouTube, the more willing we’ll be to spend money on these premium phones. Apple’s new iPhone 11 Pro and iPhone 11 Pro Max check the battery box. The phones will enjoy an additional four and five hours of battery life, respectively. How do they do that?

The answer to that question clearly illustrates the inherent advantage of Apple owning the whole stack. To learn about how that vertical integration manifests itself in a chip like the A13 Bionic, I sat down with Schiller and Anand Shimpi, who in a past life was an influential semiconductor- and systems-focused journalist who founded the website AnandTech. Shimpi is now part of Apple’s Platform Architecture team.

The new A13 outpaces last year’s A12 handsomely, with a 20 percent performance gain across all of its main components: the six CPU cores, its graphics processor, and the neural engine. For an already high-performing chip to see such a significant boost is sort of like watching Usain Bolt beat himself in a sprint.

"We talk about performance a lot publicly,” Shimpi says, “but the reality is, we view it as performance per watt. We look at it as energy efficiency, and if you build an efficient design, you also happen to build a performant design."

Shimpi and Schiller both were forceful about this maniacal focus on power efficiency and performance. For instance, the CPU team will study how applications are being used on iOS and then use the data to optimize future CPU designs. That way, when the next version of the device comes out, it will be better at doing the things that most people do on their iPhones.

"For applications that don't need the additional performance, you can run at the performance of last year's and just do it at a much lower power," Shimpi says.

This strategy isn’t just for CPUs. The same performance-per-watt rules apply to machine learning functions and graphics processing. For example, if a developer working on the iPhone's camera software sees a lot of utilization of the GPU, then she can work with a GPU architect to figure out a better way of doing things. This leads to a more efficient design for future graphics chips.

Silicon Synergy

So what happens inside the A13 Bionic when it goes to work? The general concept involves assignments, delegation, and hand-offs. For low-energy tasks—say opening and reading email—the iPhone will use the more efficient cores. But for more intense tasks like loading complex web pages, the high-performance cores take charge. For some routine and well established machine-learning work, the neural engine can hum along by itself. But for newer, more cutting-edge machine-learning models, the CPU and its specialized machine-learning accelerators lend a helping hand.

Apple’s secret, though, lies in the way all of these various parts of the chip work together in a way that conserves battery power. In a typical smartphone chip, parts of the chip are turned on to do particular tasks. Think of it as turning on the power for an entire neighborhood for them to eat dinner and watch Game of Thrones, then turning the power off, then switching on the power for another neighborhood that wants to play videogames.

With the A13, think of doing the same on-and-off approach, but on a single home basis. Fewer electrons go to waste.

"Machine learning is running during all of that, whether it's managing your battery life or optimizing performance," Schiller said. "There wasn't machine learning running 10 years ago. Now, it's always running, doing stuff."

In the end, the progression of this technology is dictated by simple things we humans want from our phones—intense games that run as smoothly on a mobile handset as a console, or a camera that takes beautiful and clean photos in the middle of the dimly lit night. As we tap and swipe, Apple’s engineers are paying attention, retooling their designs, and working on a chip for next year that will entice us to upgrade all over again.


More Great WIRED Stories