Apple Unleashes The World’s First Smartphone With A ‘3 Nm’ Processor (NASDAQ:AAPL)

Apple Inc.’s (NASDAQ:AAPL) just-announced iPhone 15 Professional would be the first smartphone to make use of Taiwan Semiconductor Manufacturing Firm’s (TSM, “TSMC”) N3 course of, sometimes called “3 nm.” The brand new course of permits a number of new options for iPhone, together with a extra highly effective graphics part with {hardware} accelerated ray tracing, and an AI accelerator that’s twice as quick because the earlier iPhone Professional technology. Though Apple received’t have the ability to maintain N3 all to itself, for now iPhone has leapt forward of the competitors in virtually each class.

Separating truth from fancy on TSMC’s N3 “3 nm” course of

A 12 months in the past, I wrote that Apple could be first to “3 nm,” and that has come to move, though the route has been longer and extra tortuous than anticipated. Additionally as I anticipated, the primary Apple gadget to function TSMC’s N3 course of is iPhone, not the Mac as some had hoped.

For nearly any new course of, it may be assumed that the primary industrial use will probably be for small space processors reminiscent of these utilized in smartphones. That is merely to enhance chip yields per wafer when the method is comparatively younger and yields are suboptimal. Beginning N3 off with the comparatively massive M-series Mac SOCs by no means actually made sense.

However earlier than I delve into the benefits conferred by the brand new course of, I have to take just a few paragraphs to dispel some myths about N3, which sadly, Apple itself is now responsible of promulgating. Within the September 12 occasion video, Sribalan Santhanam, VP of Silicon Engineering comes out and says “parts of those transistors are simply over 12 silicon atoms large.” That’s a minimum of very deceptive.

With out understanding what “parts” Santhanam was referring to, I can’t say completely that he’s fallacious. In some areas, deposition layer thicknesses could also be that skinny, however that’s not likely vital from the standpoint of transistor areal density, which is what number of transistors you possibly can placed on the flat floor of the chip, measured in hundreds of thousands of transistors per sq. millimeter.

Producers characterize how tightly transistors may be packed aspect by aspect on a wafer by just a few key dimensional parameters, such because the spacing between the steel contacts of adjoining transistors, known as the Contacted Poly Pitch or CPP.

On the Worldwide Electron Gadgets Assembly in 2022, TSMC introduced a paper on its N3 course of, and disclosed the CPP for N3. This has been summarized by Scotten Jones in SemiWiki in a desk proven beneath:

TSMC didn’t reveal any of the opposite attribute dimensions proven within the desk, which is why they’re clean for N3. However even scaled proportionally by the change in CPP, none of them would method the 12 silicon atom width, which is about 1.6 nm.

And that is one thing that Intel (INTC) has been complaining about for a while, justifiably. There’s actually nothing by way of significant bodily transistor dimensions that corresponds to three nm, not to mention some fraction of it.

Producers like TSMC and Intel have been transferring away from calling out nanometers. TSMC calls its processes, N7, N5, N3, and Intel calls its processes Intel 7, Intel 4, Intel 3, and so on. However it nonetheless makes for good advertising copy for foundry prospects like Apple to check with nanometers.

We don’t but know what transistor density Apple was capable of receive for the A17, however TSMC has been claiming that N3 would supply a 70% improve in transistor density in comparison with N5. Apple’s first N5 processor was the A14 Bionic, which had a complete transistor depend of 11.8 billion on a silicon die dimension of 88 mm^2 for a sensible density of 134 million transistors per mm^2.

The brand new N3 course of will possible improve transistor density to about 227 million transistors per mm^2. For the primary technology N3 processors, Apple has possible opted for a chip dimension discount with a view to enhance yield, whereas nonetheless growing transistor depend. Apple claims the brand new A17 Professional has 19 billion transistors, for a possible chip dimension of about 84 mm^2.

The A17 Professional has greater than 50% higher transistors per chip than A14. Final 12 months’s A16 Bionic had “practically 16 billion transistors,” so the A17 Professional has a minimum of 3 billion greater than A16. 3 billion transistors interprets into loads of extra circuitry and functionality. In the remainder of this text I’ll take a look at how Apple has chosen to make use of this new functionality.

How Apple has used the three billion additional transistors of the A17

The A17’s transistor depend is just about 20% higher than A16, so Apple needed to be cautious the way it spent the brand new transistors. Apple apparently didn’t lavish many new transistors on the CPU, which is unchanged in core depend and solely about 10% sooner than the A16.

This is sensible, for the reason that A16’s CPU efficiency was already higher than the closest competitors, the Qualcomm (QCOM) Snapdragon 8 Gen 2, which may be discovered within the Samsung (OTCPK:SSNLF) Galaxy S23 Extremely. Though it’s solely one among many benchmarks, in Geekbench CPU testing, A16 mops the ground with the Snapdragon:

Click on to enlarge

Each the A16 and Snapdragon are fabricated on TSMC’s N4 course of, so the Snapdragon wasn’t deprived by way of course of. Apple’s customized ARM CPU cores proceed to guide the business in efficiency and energy effectivity.

The place Apple did spend its additional transistors may be seen from this abstract of A17 enhancements:

Apple fully revamped the GPU with a brand new shader structure and added an extra GPU core in comparison with A16. The GPU now options {hardware} accelerated ray tracing, a primary for Apple.

Though the GPU acquired an enormous enhance, I think that Apple saved most of its transistors for the brand new A17 Neural Engine. Though it has the identical variety of cores because the A16, it’s twice as quick.

New options enabled by A17: higher gaming with playable ray tracing, and transformer AI on gadget

What do these new {hardware} options imply for customers? The most recent technology of recreation consoles have featured {hardware} accelerated ray tracing for years. For the reason that creation of those consoles, Apple has not been ready credibly to say “console degree efficiency” for iPhone, iPad, and even Mac.

With {hardware} accelerated ray tracing, which is augmented by the Neural Engine, Apple can start to meet up with console cutting-edge and entice extra recreation builders to the platform. Ubisoft is there with The Division Resurgence:

Resident Evil Village and Resident Evil 4 (remake) may even be on the platform. The most recent model of the Murderer’s Creed franchise, Mirage, can be coming to iPhone.

Revitalizing gaming on Apple platforms is significant to increasing the constituency for Apple Silicon. One of many main shortcomings that reviewers discover for Apple computing units is lack of help for widespread PC and console video games.

Moreover, efficiency for Rosetta-converted Intel-native video games on the Mac has not been superb. Now that Apple has totally transformed its Mac lineup to Apple Silicon, all of Apple’s computing merchandise will have the ability to construct on the A17’s graphics.

Apple appears to be dedicated to on-device AI, and that now contains pre-trained transformer fashions. Right here, Apple is swimming in opposition to the prevailing present, which primarily locations generative pre-trained transformers within the cloud due to their very demanding computational necessities. However transformers may also run on native units if the scope of their capabilities is sufficiently slim.

Apple now makes use of a transformer hosted within the Neural Engine as the idea for Siri speech recognition and textual content completion in messaging. There’s hardly any operate of iPhone that isn’t enhanced by AI in a roundabout way.

Computational pictures is AI primarily based and gives a brand new operate to mechanically discover individuals in a picture to help portrait mode. Computational pictures additionally helps capturing what Apple calls “spatial video,” or stereo video utilizing the principle 48 MP digital camera in addition to the 12 MP ultrawide digital camera. These movies can then be considered in stereoscopic 3D on Imaginative and prescient Professional.

Even plain previous voice calling will get a lift with AI-based background noise suppression. Background noise suppression is definitely a tough downside whenever you don’t have a separate background audio sign freed from consumer’s voice. However it’s a superb downside for AI, and Apple’s method might be the best obtainable.

Investor takeaways

Now that each Apple platform relies on Apple Silicon, it’s turn out to be extra vital than ever for traders to concentrate to the iPhone launch. The place Apple goes with the A sequence SOCs, the remainder of the product line will comply with.

A very good instance is the brand new Sequence 9 Watch, which acquired the brand new S9 System in Bundle (SiP). For the primary time, even the little S9 acquired a Neural Engine, and transformer primarily based Siri dictation together with it.

The emphasis on graphics enhancement in A17 Professional is as a lot a response to what’s taking place within the Mac product line as it’s to iPhone itself. On the excessive finish of the road, the place the M2 Extremely is used, CPU efficiency could be very spectacular, however GPU efficiency, not a lot.

Within the realm of excessive finish graphics workstations, Apple actually can’t compete with PCs outfitted with Nvidia’s (NVDA) RTX 4090. No less than not but. However the A17 gives a superb basis on which to construct a way more succesful GPU for iPads and Macs.

And because the world rushes to embrace generative pre-trained transformers (GPTs), Apple knew it wanted to enormously improve its on chip AI if it had any hope of maintaining. In fact, Apple can’t run a ChatGPT on an iPhone, however it may nonetheless make use of transformer fashions to carry out helpful capabilities and has began to take action with A17.

We’ll see much more succesful Neural Engines primarily based on the identical structure in future Macs. For the Mac, a scaled down ChatGPT will not be inconceivable, maybe as a Siri alternative (however nonetheless known as Siri). This might be crucial for Apple.

GPTs principally function clever engines like google and can in all probability exchange the normal search engine over time. Apple hasn’t been capable of be aggressive in search and has remained depending on Google (GOOG). With the arrival of GPTs, that would change. And it’s actually conceivable that Apple might present its personal cloud-based GPT service solely for Apple units, which might work at the side of on gadget AI.

General, iPhone 15 Professional is a really succesful little laptop, which may be related to a 4K monitor by way of USB-C. Future iPads will use the A17 or processors primarily based on it, and these will probably be full-fledged computer systems in their very own proper with keyboards, mice and exterior screens related by way of USB-C.

iPhone 15 and 15 Professional will do very properly within the market, and should even revive the smartphone business, which has been within the doldrums. I search for iPhone to renew income development in fiscal 2024. I stay lengthy Apple and charge it a Purchase.