As reported by Bloomberg, citing its own informants from the depths of Apple, the Cupertino company is going to use its own processors in Mac computers starting in 2020 instead of Intel chips. The initiative is codenamed Kalamata (after the Greek city of Kalamata?) and is still in the early stages of development.
However, Kalamata is said to be part of a larger strategy that will bring all of Apple's devices, including Macs, iPhones and iPads, closer together, both in terms of hardware and environment. The project, approved by Apple management, will most likely lead to a phased transition. However, as always, even if this information is accurate, Apple may still cancel or postpone plans.
The moves could be a major blow to Intel, whose partnership has helped revive Apple's success in the desktop market while also working to improve the image of one of the world's largest chip makers. Apple's contracts account for about 5% of Intel's annual revenue, according to Bloomberg. Following the spread of persistent rumors, Intel shares fell by 9.2%. This is the biggest one-day decline in the company's stock price in more than two years.
For Apple, this transition promises to be a defining moment. Currently all iPhones, iPads, Apple Watch and Apple TV use single-chip systems developed within the depths of Apple and based on the architecture of ARM Holdings. The transition to its own chips in the Mac will allow Apple to release new models at a time that is optimal for itself, rather than relying on the Intel product update cycle.
This will also allow the Cupertino company to quickly introduce new features across all its products and differentiate them from competitors. Using its own SoCs would make Apple the only major PC maker that doesn't rely on third-party processors. Dell, HP, Lenovo and ASUS use x86 chips from Intel and AMD. Apple will also be able to more closely integrate new hardware and software, achieve improved battery life and other benefits.
And although the transition to Apple chips in the hardware sector is planned to begin only in 2020, changes on the software side will begin earlier. iPhones and iPads with Apple chips use the iOS operating system, while Mac computers run macOS. Over the past few years, Apple has gradually brought both platforms closer in terms of user experience and environment, and more recently has begun to look at porting lower-level features such as new system file management.
As part of a larger effort to make the Mac work more like the iPhone, Apple is working on a new software platform, called Marzipan (marzipan is a confectionery product). It could be released as early as this year and is said to make it possible to run iPad apps on Mac computers.
Apple has begun integrating ARM coprocessors into its computers, which run iOS-like code to implement features like security or controls. The company's latest laptops and all-in-one PCs received such coprocessors. It is expected that the next version of the Mac Pro 2019 will receive them. Also, according to informants, Apple single-chip systems will be used in the company's new laptops, which will be released this year.
Intel has long dominated the processor space, increasingly taking share away from Advanced Micro Devices (AMD), its only competitor in the x86 market. Intel also makes modems for some iPhone models so they can connect to cellular networks. Although Apple lags behind many companies in terms of computer shipments, it is in the top three in terms of revenue.
In 2005, Apple announced a transition from IBM PowerPC to Intel x86 chips in its Macs, an initiative announced on stage by former Intel executive Paul Otellini along with the late Apple co-founder Steve Jobs. Such transitions between architectures are always very painful and require a lot of effort on the part of both OS owners and software developers. If effective emulation methods are not developed, the latter will have to rewrite the code and recompile their x86 programs for the ARM architecture. , the task of effective emulation is quite non-trivial and, in addition to compatibility issues, acutely raises performance issues. However, this can be somewhat compensated for by the appearance of support for iOS applications in macOS (but this immediately raises the question of the lack of touch screens on Mac).
Another problem with the transition is that until recently, all of Apple's ARM chips were optimized primarily for lightweight mobile devices. Therefore, it is logical to assume that the transition will first affect thin laptops, and then Apple will gradually move higher, trying not to sacrifice the performance and capabilities of its Macs. However, rumors and speculation about the transition of Apple Mac to ARM are not new and arise from time to time. For example, in 2014, one of the former Apple executives, Jean-Louis Gassée, said that the first ARM-based Macs would appear in 2016.
After Apple used its own A4 processor in the first iPad, rumors began to spread that in the future the company might abandon Intel processors in Macs and switch to ARM architecture. There are benefits to this, but there are a lot of implications that this migration will have that Apple will need to overcome. Is the game worth the candle?
Why modern Macs run on Intel processors
Since 2006, all new Apple computers run on x86 processors in conjunction with a GPU from Nvidia or AMD (or integrated Intel graphics in entry-level models). Thanks to Open GL, Mac software can interact with GPUs different architecture, allowing Apple to easily change graphics providers.
After the transition to Intel processors, Apple released two more major releases of OS X, which included support for the old (PowerPC) and new (x86) architecture, but Snow Leopard, released in 2009, only worked on Macs with Intel processors.
How is switching to ARM different from switching to Intel with PowerPC?
In the period from 1994 to 2005, all software for Mac OS was “tailored” exclusively to work on PowerPC processors, the architecture of which was radically different from x86. Even earlier, for the previous ten years, Macs were running Motorola processors called 68k (68000, 68020, 60030 and 68040).
The first change in architecture was caused by the desire to move to more modern and productive processors with support for 64-bit computing. With PowerPC's performance far superior to 68k, it could easily emulate existing code.
Apple's second transition, from PowerPC to Intel, didn't look like such a big step forward. Manufacturers of PowerPC chips (IBM and Motorola/Freescale) have actually left the PC market, “playing cameo roles” in the niches of the automotive industry and game consoles. Apple was their last client, but the company sold less than 4 million of its computers a year.
But the Windows PC market was in full swing, all computers used Intel x86 architecture or compatible analogues from AMD. By switching from PowerPC to Intel, Apple abandoned a sinking ship and chose an emerging ecosystem where innovation and technology developed very quickly due to high volumes of production.
However, the available x86 architecture was essentially a step backwards. After all, at that time all Intel processors were 32-bit, while the PowerPC, which Apple used in its PowerMac G5 since 2003, supported 64-bit computing. It wasn't until 2006, when Intel introduced the Core 2 line, that Apple returned to 64-bit processors in its computers.
There were other shortcomings in the transition to Intel architecture, but they were covered by the high pace of development due to the large market. At that time, Intel processors were slightly more powerful than PowerPC, but their performance was sufficient to emulate most code written for PowerPC. This was made possible thanks to Rosetta technology, which Apple bought and improved to smooth out the difficulties of moving to a new platform.
In addition, changing the architecture to x86 meant the ability to run Windows (Linux and other x86 OS). This significantly expanded the potential audience, attracting users who needed to run specific Windows applications to purchase a Mac. Boot Camp allowed Windows to be installed as a second system on disk, and third-party applications made it possible to run Windows programs directly within OS X. Both methods were significantly faster than simply emulating Windows code on PowerPC, which was the only option available to Mac users before the transition to Intel processors .
Why Apple might be interested in moving away from Intel processors
Saving money
The main reason why Apple might consider making Macs without Intel processors is the high price of the latter. Intel's chips are too high-tech and difficult enough to copy, which is why they are uncompetitive and allow Intel to charge such a high price for them.
It's difficult to determine the exact price Apple pays for Intel processors. Analysts from IHS iSuppli estimate that the Intel Core i5 used in the Microsoft Surface Pro is 4-5 times more expensive than the ARM chips in the Surface RT. A6 processors for iPad, in their opinion, cost Apple $25 apiece, while Intel chips used in Macs cost $180–$300. The idea that Apple could replace Intel's $200 chips with one or two $25 chips sparked rumors about the possibility of Apple computers switching to ARM architecture.
However, such a comparison is not entirely correct, because modern ARM processors are significantly inferior in performance to even entry-level Intel Core i5 chips. There is a huge gap between the processing power of Intel processors and the fastest ARM processors - this was proven by Microsoft's experiment in porting Windows to the Surface RT ARM chip.
Apple can create more powerful ARM processors
Apple has been aggressively increasing the processing power of its Ax series processors, thanks to funding from production savings. The company sells about 70 million iPads and nearly 170 million iPhones each year.
Apple could have created even more powerful A8 chips this year if not for the limitations imposed by case thickness, limited battery size and heat dissipation issues in iOS devices. The company has made it clear that when designing the A8, the top priority was energy efficiency, which is so important for the iPad Air 2 (which has a smaller battery than its predecessor), in order to maintain the same battery life.
Mac mini and even MacBook Air much less constrained by power consumption and heat dissipation limitations, which would allow Apple to increase processor operating frequencies, the number of cores, or add other hardware to them, providing them with larger amounts of memory and cache.
Given all these circumstances, Apple may even be interested in creating some new specific Mac running on an ARM processor, which in terms of performance will not be so far from budget desktop processors. After all, ARM is now already ahead of Intel x86 mobile chips.
Before switching to Intel processors, Apple produced about 4 million Macs a year. On this moment Annual Mac production volumes are almost 20 million, and the company sold about the same number of iPads in the first four quarters. Apple initially considered using Intel Atom chips in the iPad, but abandoned the idea in favor of ARM.
Creating our own silicon IC technology
Based on the fact that Apple uses optimization of chips used in iOS devices, we can assume that the company is also interested in optimizing processors for Mac. It can remove unused logic and introduce additional ones to implement encryption, audio processing, or video decoding in hardware.
Using a single architecture across Macs and iOS devices can greatly simplify the use of hardware, software, and API migration and other software between systems.
Moreover, when developing proprietary technologies used only in Ax processors, all Apple investments will remain within the company and will bring profit exclusively to it. Now, by buying processors from Intel, Apple indirectly makes a contribution to the development of the entire PC industry. Intel is creating new generations of processors that are available to everyone, and their development costs are reduced by the production volumes provided by Apple.
Given Intel's lackluster success in motivating PC makers to create Ultrabooks, Mac mini clones, and Android Atom tablets, losing a customer like Apple would have catastrophic consequences not only for Intel, but for everyone who uses x86 processors.
What's holding Apple back from switching to ARM
Apple made a move towards Intel for reasonable reasons. In 2006, it did not have a serious team for developing chips, nor did it have sufficient capital for development own technology their creation. Intel had already done the job, and buying an off-the-shelf solution not only made sense, but was the best of the few options available to Apple at the time.
Despite the fact that Apple is now one of the leading manufacturers of mobile processors and has $150 billion in capital, allowing it to implement the most ambitious projects, using Intel chips still makes sense for a number of reasons.
Existing Intel Technologies and Capabilities
Today, Intel has the world's leading processor technology and amazing manufacturing capabilities to match Apple's demands. By remaining an Intel customer, Apple not only gets them, but also the chipmaker's future designs, which it will invest in to remain the most advanced processor manufacturer in the world.
Large orders give Apple priority in chip selection, as well as discounts due to high volumes. The profit the company receives from each Mac sold is simply unattainable for PC manufacturers, even taking into account the considerable cost of Intel processors.
For Apple, there are no half measures that other manufacturers can take; it chooses only the most advanced technologies. The company buys the best LCD panels and uses the licensed Helvetica font. While Microsoft and Google use low-quality displays, copies of Helvetica, and also do not use fingerprint scanners in their products due to their high cost.
Loss of AMD as a supplier
By leaving Intel, Apple may lose a potential supplier of x86 compatible video chips from AMD.
The company now purchases GPUs from both AMD and Nvidia, choosing the best from available solutions depending on new technologies and price. Thanks to OpenGL, changing video chip vendors is easy.
Apple did not play into AMD's hands in their confrontation with Intel, but theoretically it could - if Intel makes a mistake and AMD manages to create a more affordable and superior processor capable of running x86 code on Macs. Apple's move away from Intel to ARM processors will eliminate even this theoretical possibility of replacing Intel chips with cheaper AMD ones.
Dubious savings when partially switching to ARM
Apple now will not be able to replace Intel processors with ARM in the entire line of Macs, especially in the top families and modifications MacBook Pro and Mac Pro, and this is precisely the segment from which the company receives most of its profits and, thanks to minimal competition, retains the loyalty of the community.
If Apple releases just one new Mac model running on the ARM architecture, it will reduce its dependence on Intel, but will also increase the cost of purchasing processors for x86 Macs by reducing volumes. Thus, a partial transition to ARM will not give Apple anything in terms of savings.
The mere fact of creating an ARM Mac does not guarantee its popularity. Microsoft has already made an attempt to port Windows to ARM, but this did not attract a new audience. Two years were wasted, except for the deterioration of relations with Intel. The processor giant responded by announcing support for Android and Meego/Tizen, spending billions of dollars subsidized by tablet manufacturers to introduce Atom, which was aimed at the same goal as Microsoft with its Surface RT - significantly expanding the market.
Of course, Microsoft was not going to save money and the main reason for using ARM was the desire to increase energy efficiency compared to desktop and mobile alternatives from Intel. But these wonderful initiatives were killed at the root by the harsh reality - existing Windows applications could not run on ARM architecture.
Apple has great experience in porting software to new architectures. The company has proven that it can simultaneously support multiple hardware platforms, but despite this, it has always tried to complete such transitions quickly to bring everything to unified standard and avoid the problem of hardware fragmentation.
Big risks
On top of the financial side of things, developing ARM chips for the Mac could pose additional problems, such as complications and slowing down the development of mobile processors used in iPhones, iPads and other new products.
Apple's mobile device sales make up the majority of its profits. Last year the company sold 244 million iOS devices and only 18.9 million Macs. The transition to ARM architecture will inevitably cause a change in priorities for the development of the mobile segment and, theoretically, could allow competitors to become leaders. It is unlikely that Apple has hundreds of free engineers sitting idle to disperse the efforts of the ARM chip development team into two different directions.
By distancing itself from a key supplier, Apple could also confuse existing customers and risk tarnishing its name. When Microsoft introduced the Surface RT, it lost customer trust because the "no-compromise Windows PC" couldn't actually run Windows apps and was limited by the performance of ARM processors. Potential buyers of ARM Macs will have even greater demands and expectations from the new Apple product.
Incompatibility with x86 architecture
Apple has extensive experience in porting its own operating systems, frameworks, applications and development tools to new architectures. The company transferred Mac OS from 68k to PowerPC, ported NeXT software from Intel to PowerPC, and iOS, in fact, is OS X adapted to the mobile realities.
Apple certainly knows how to build an ARM version of OS X and, if necessary, can provide tools for developers to help them rebuild their applications for Macs on ARM architecture, but this will require a lot of work and significant effort from the developers themselves. The costs and expenses associated with creating application ports may not meet expectations, especially if Apple sells less than 20 million Macs per year.
Apple TV experience
Like Surface RT, Apple TV can be seen as a recent example of a shift in architecture. The original Apple TV, sold from 2007 to 2009, was essentially a stripped-down Mac with an Intel x86 processor and Nvidia graphics, running a modified version of OS X.
In 2010, Apple introduced the second generation of TV set-top box, running iOS on its own A4 processor, which had built-in graphics. This transition, which entailed a complete redesign of the hardware architecture, reduced the price of the product from $299 to $99.
But Apple TV is very specific example- the set-top box is produced in relatively small volumes and does not bring much profit to the company; moreover, it does not have third-party applications, which means there are no problems with their adaptation. Its transition to iOS and ARM was a fairly simple task. At $300, the Apple TV simply didn't have a chance in the market, but when it dropped to $99, the set-top box began to sell very well, bringing in nearly a billion dollars a year for Apple (including media content that boosts its sales). In 2010, Apple had a source of rejected A4 (and then A5) chips that were not suitable for the iPad, so Apple TV became an ideal candidate for the transition to ARM architecture.
Don't expect an ARM-MacBook in the near future
The question of traditional Macs switching to ARM architecture is not whether Apple can replace Intel, but rather whether it will be commercially viable.
If Apple really decides to introduce an ultra-low-cost MacBook Air in a “netbook format,” then it will be easier for it to abandon expensive Core i5 chips and create an inexpensive product that runs on iOS or a stripped-down version of OS X. Such a MacBook would take its place next to the Surface RT and Chromebooks. from HP and Samsung, running on ARM chips from Samsung.
However, at the moment there is too little convincing evidence that would prove that Apple is interested in selling laptops with low performance. A record number of Macs are now on sale. price category$900-$3000, and there is also an iPad that covers the more budget range of $200-$800.
Despite the fact that iPad sales decreased by 4% at the end of last year, it cannot be said that the tablet form factor is losing popularity and needs to be replaced. In fact, it looks like Apple has turned iPad users into potential Mac buyers, which is a much bigger success (and profit) than motivating Mac users to buy an iPad.
Yet the technology industry is constantly in flux, and routines are often disrupted by new products that cost and do less than existing ones. Proof of this can be seen in the iPhone, which was capable of significantly less than the existing smartphones at that time, as well as the iPad and Apple TV, which lacked the functions that were in the TV set-top boxes that preceded them. Apple simply cut off the “necessary” features and thereby created new, accessible and attractive product categories.
By creating a Mac based on an ARM processor, Apple could greatly discredit own business premium computers. Theoretically, the company could create an inexpensive MacBook for, say, the education sector, but this is too small a market that is now saturated with Google Chromebooks.
In a year or two, circumstances may change. It's possible that Apple will reach a point where its premium Mac business will be difficult to expand further. During this time, the company can develop technology that would allow it to create an ARM processor that is very close in performance to Intel, but has more low price. Apple could create hardware support for emulating x86 applications, thereby minimizing costs and speeding up the transition to ARM.
While Intel doesn't have any major breakthroughs in x86 processor development, it may make more sense for Apple to invest in the design and development of its own modern ARM chips (or even a completely new architecture) for desktops and laptops.
In general, it seems that the market for traditional computers and laptops has stopped developing. Apple is expanding its share of premium computers, and it has every chance to continue this trend without making radical changes to the Mac. The company can use its huge, but still limited resources for more profitable investment than replacing Intel as the processor supplier for several million Macs. At least for the next few years.
Why is iPhone 7 faster than Samsung Galaxy S7, and iPhone 8 faster than Galaxy S8? The point here is the different ideology of operating systems, and besides, one of the main advantages of Apple was and remains unique systems on a chip. The A10 and A11 processors are noticeably ahead in benchmarks similar offers from Qualcomm represented by Snapdragon 820/821 and Snapdragon 835, respectively. Why is this happening? What is the “Apple magic”? Leaving aside arguments in the style of “Android is better!”, Let's try to understand the reasons that led to the superiority of Apple mobile processors over Qualcomm offerings.
Factor one: it just happened
Let's remember 2013. Qualcomm has very successful Snapdragon 800 chips in its arsenal, based on 32-bit Krait 400 cores of its own design. Dozens, if not hundreds of a wide variety of models were released on this chip (and its successor, Snapdragon 801). At the time of its announcement, Qualcomm’s top-end chipset simply had no alternatives: solutions based on ARM Cortex A15 cores were extremely power-hungry and could not compete with four custom Krait cores. Everything seems to be fine, Qualcomm is the king of the hill, it’s enough to continue to develop successful architecture. It would seem, what could go wrong?
But - in order. In 2011, ARM Holdings announced the ARMv8 architecture, the use of which opened up numerous possibilities for accelerating parts special types computing - for example, stream encryption, which (let me get ahead of myself) is used in almost all smartphones today. The first mobile cores of this architecture were Cortex A53 and A57, announced by the ARM holding in 2012. At the same time, ARM predicted the release of finished processors on new cores only in 2014. But Apple, the holders of the ARM architectural license, was the first to do so - almost a year earlier than its competitors.
So, in November 2013, Apple released the iPhone 5s. In addition to a fingerprint sensor and built-in Secure Enclave security, the new iPhone features a market-first Apple A7 64-bit ARMv8 processor. The new processor shows miracles of performance in Geekbench: the result of a dual-core processor in single-threaded calculations is one and a half times higher than the results of the Krait 400 cores, and parity is observed in multi-threaded ones.
The expanded set of ARMv8 commands could not have come at a better time: it was in the iPhone 5s that Apple built in the Secure Enclave hardware security system, which is also responsible for data encryption. From Apple's point of view, the choice of 64-bit architecture was quite logical: only in kernels with ARMv8 support did instructions appear to speed up stream encryption, which at that time had been used by Apple for quite some time. Subsequently, the use of new cores allowed Apple to achieve unprecedented access speeds to encrypted data - the Nexus 6, released a year later, based on the 32-bit Qualcomm Snapdragon 805 (ARMv7), showed terrifying crypto streaming performance: encrypted data was accessed 3-5 times slower than unencrypted ones.
At first, the 64-bit architecture in smartphones was perceived by ordinary people - and many experts - as pure water marketing. This is what users thought, and this is what Qualcomm executives said - at least in their official speeches.
In 2014, the iPhone 6 was released, equipped with an A8 processor, also working with the ARMv8 instruction system. How does Qualcomm respond? Small update: the market is dominated by smartphones running on Snapdragon 801 (32 bit, ARMv7). Also coming out is the Snapdragon 805, which uses the same Krait 400 cores but with a more powerful GPU. Apple processors turn out to be faster than their Qualcomm counterparts in both single-threaded and multi-threaded calculations, and in specific applications - for example, in the implementation of stream encryption - they simply outperform competitors' solutions by several times. Qualcomm is trying hard to pretend that nothing unusual is happening, but manufacturers are stepping on the throat and demanding a competitive SoC. Qualcomm has no choice but to join the race.
In 2015, Apple released the iPhone 6s and A8, Qualcomm released the Snapdragon 810 chip and its stripped-down version, the Snapdragon 808. These processors were Qualcomm’s response to the demands of its partners. However, the lack of experience in developing 64-bit chips played a cruel joke on the company: both processors turned out to be extremely unsuccessful. From the very first days, the processors showed a tendency to excessive power consumption, overheating and throttling, as a result of which their established performance after a few minutes of operation differed little from the performance of the Snapdragon 801.
What conclusion can be drawn from all this? There is only one conclusion: Apple took the industry by surprise by using cores with a new architecture when and where there seemed to be no need for it. As a result, Qualcomm found itself in the role of catching up, and Apple received a head start of a year and a half. Why did it happen?
Here we need to consider the features of the development cycle of mobile processors.
Factor two: difference in development cycles
So, we found out that Apple managed to get ahead, a year and a half ahead of its competitors. How could this happen? The reason is the difference in development cycles between Apple and Android smartphone manufacturers.
As you know, Apple has complete control over the development and production of the iPhone, starting from the lowest level - processor design. And if until recently Apple licensed graphics cores from Imagination Technologies, the company preferred to develop processor cores independently.
What does Apple's development cycle look like? Based on the ARM architectural license, a processor compatible with a given instruction set (ARMv8) is designed. At the same time, a smartphone is being developed that will use this processor. In parallel, all the necessary drivers and OS are created for it, and optimization is carried out. Everything happens within one company; OS developers have no problem gaining access to driver source code, and driver developers, in turn, have the opportunity to communicate with the people who designed the processor.
Production cycle Android devices look completely different.
First of all, ARM, the developer of instruction systems and processor architectures of the same name, comes into play. It is ARM that designs reference processor cores. Thus, back in 2012, ARM Cortex A53 cores were announced, on which the vast majority of smartphones released in 2015, 2016 and 2017 are based.
Wait a minute! 2012? That's right: 64-bit A53 cores were announced in October 2012. But the core architecture is one thing, and real processors are quite another: ARM Holdings simply does not produce them, offering reference designs to partners, but not supplying the SoCs themselves to the market. Before a smartphone based on a particular architecture appears on the market, someone must develop and release a complete system on a chip, SoC.
Despite public performance own representatives, in 2013 Qualcomm worked hard to release a 64-bit processor. There was no time left to develop our own kernel; I had to take what they gave me. They gave us the big.LITTLE architecture, which at that time included “small” Cortex A53 cores (successful) and “large” A57 cores (quite controversial from the point of view of energy efficiency and throttling).
The first Qualcomm processors based on these cores were announced in 2014. But the processor is not everything! At a minimum, you also need a case, a screen... All this is produced by OEM manufacturers, who, in fact, are engaged in the development and production of smartphones. And this is also time, and considerable time.
Finally, operating system. In order to run Android on a device, you need a set of drivers for the new chipset. The drivers are developed by the chipset developer (for example, Qualcomm), providing them to smartphone manufacturers for integration. It also takes the manufacturer some time to understand and integrate the drivers.
But this is not the end! A ready-made smartphone with a working version of Android must also be certified in one of Google’s laboratories for compatibility and compliance with the Android Compatibility Definition. This is also time, which is already catastrophically short.
In other words, the fact that we saw smartphones with Snapdragon 808/810 only in 2015 is not at all surprising. Qualcomm's first flagship chips, based on 64-bit architecture, were a year and a half behind Apple's SoC. This historical fact, and this is Apple's real advantage.
In 2015, the long development cycle and the requirements of partners played a cruel joke on Qualcomm: the first pancake turned out to be lumpy. However, the company managed to redeem itself with the release of Snapdragon 820. But was it too late?
Factor three: the issue of size
Let's look at a table that compares the two latest generations of processors from Apple and Qualcomm.
What do we see from this table? It is easy to see that the performance per core in Apple processors is more than two times higher than Qualcomm solutions, and the multi-threaded performance of current generations of processors differs by almost one and a half times. Why does this happen? You can try to find the answer in the following table.
If we put aside the A10 Fusion/Snapdragon 820 pair of processors, which use different process technologies, we can compare the surface area of the A11 Bionic and Snapdragon 835 chips. Apple's chip has 1.2 times the surface area of Qualcomm's solution. What does this mean in practice? The ability to use more transistors, more advanced core architecture. In particular, the researchers found that the A11 Bionic's "weak" processor cores are several times larger than the small A53 cores (sorry - Kryo 280) used in the Snapdragon 835. This means that even the "small" A11 Bionic cores support out-of-order execution of commands. which allows you to get greater performance per clock compared to straight A53 cores.
The area of the processor directly affects its price. The larger the area (when using one technical process), the higher the cost. Which brings us to the next factor: the cost of the processor for the manufacturer.
Factor four: the issue of price
According to a report by Android Authority, the Apple A10 Fusion's processor core area is twice that of its closest competitor, the Snapdragon 820.
“Apple’s advantage is that the company can afford to spend money on increasing the area of the processor, built using the latest 16-nanometer FinFET technology... A few extra dollars will not play a big role in the final cost of the device - but Apple will be able to sell significantly more than 600- dollar devices due to such great performance,” writes Linley Gwennap, director of The Linley Group.
Indeed, an extra five or six dollars will not play a big role in the final cost of the iPhone - these are fractions, in the worst case, a few percent of its cost for the consumer. But if those five or six dollars can double the device's performance compared to Android competitors, that's a great argument in Apple's favor.
Why doesn't it work out for Qualcomm? There are too many stakeholders in the development chain of processors for Android devices. These include ARM, which develops and licenses processor cores, and Qualcomm, which designs ready-made processors under license, and manufacturers of Android smartphones. With OEMs forced to compete on price, every dollar counts. Manufacturers want the cheapest possible SoCs (which is why, by the way, solutions built on the archaic weak A53 cores are still so popular), and Qualcomm has to take this into account. But both Qualcomm and ARM want a piece of the pie, getting their share of the profits, so the cost of a solution similar to Apple processors would be even higher than Apple's. As a result, OEMs would not be able to afford bulk purchases of such processors, which would further increase their cost. (By the way, this is exactly what happened with the MTK Helio X30 processor - it was not in demand, and only two smartphones were released based on it.)
Of course, an argument can be made here that Samsung and Huawei have their own processor lines - Exynos and Kirin, respectively. But Huawei does not have its own developments; the company takes ready-made ARM Cortex cores and ready-made ARM Mali graphics accelerators, assembling “its own” processors based on them. It is clear that the computing cores of these processors cannot be more powerful than those offered by ARM. Samsung is trying to follow Apple's path, releasing its own customized cores - the performance of which, however, is not far from the usual "stock" ARM cores.
Factor five: the issue of control
Last year, Apple did an interesting thing: by a willful decision they removed support for 32-bit applications from iOS 11. Coincidentally, it was on this version of the OS that the new iPhone line was released: 8, 8 Plus and X. What does this mean in terms of performance ?
The ability to take and abandon support for 32-bit commands gives a lot, a lot. The decoding and execution blocks are simplified, and the required number of transistors is reduced. Where does it go this saving? It can be spent on reducing the processor area (which directly translates into reduced cost and reduced power consumption), or you can, with the same area and power consumption, add transistors to other blocks, thereby increasing performance. Most likely, it was the second scenario that events unfolded and the A11 Bionic processor gained an additional 10–15% of performance precisely due to the refusal to support 32-bit code.
Is this possible in the Android world? Yes, it’s possible, but not completely and very soon. Only from August 2019, requirements for developers who will have to add or update applications to Google Play Store is required to include 64-bit versions of binary libraries. (We note here that not all - and not even the majority! - Android applications generally use any binary libraries, often content with dynamically translated bytecode.) Let us recall that Apple introduced a similar requirement in February 2015 - again a time advantage, this time of four and a half years.
Factor six: optimization and use of available resources
Optimization is a critical component of productivity. Traditionally, with Apple, everything was either perfect or exemplary with optimization (users who complain about decreased performance of old devices updated to latest version iOS, they simply don’t understand what hell it would be like on such weak hardware if they ran Android on it). But with Android optimization everything is... motley. Diverse. One might say - enchanting.
Most often, pure builds of Android - such as those used in Google Nexus and Pixel smartphones, Motorola and Nokia devices - work quite quickly on fresh hardware. But even here, not all is well: for example, in the Google (Motorola) Nexus 6 smartphone there were absolutely stunning problems with the speed of access to the drive, which arose due to an illiterate implementation of encryption (Google developers could not cope with the hardware accelerator of cryptographic operations Snapdragon processor 805, after which they stated that “software implementation is better”). In this article, we analyzed in detail the speed of reading and writing encrypted data by the Nexus 6 smartphone, comparing it with the speed of similar operations in the iPhone 5s. Here are the numbers:
- Nexus 6, sequential read, unencrypted data: 131.65 MB/s;
- Nexus 6, sequential read, encrypted data: 25.17 MB/s (39 MB/s with Android 7 upgrade);
- iPhone 5s, sequential read, encrypted data: 183 MB/s.
Impressive? With similar hardware characteristics, Google developers (Google, not the crooked OEM!) managed to make such a mistake in a reference device that was supposed to promote secure encryption to the masses. Would you be surprised to learn that other manufacturers may have problems with optimization? And they arise. Thus, the fully equipped HTC U Ultra (Snapdragon 821) manages to slow down and overheat during the most routine operations; it feels like the processor is doing at least twice as many calculations as it should. Well oh Samsung smartphones, which manage to slow down in small ways even on the most powerful hardware available, is not even worth talking about in detail.
Factor seven: screen resolution
There is one more point that is worth mentioning. This is the display resolution. As you know, standard iPhone models are equipped with screens with HD resolution, Plus models - Full HD. Manufacturers of smartphones running Android, using flagship Qualcomm chipsets, try to install screens with a resolution of QHD - 2560 × 1440. Well, at the very least - Full HD, but this is, alas, rare in flagship smartphones.
Why "alas"? Because resolutions higher than Full HD on screens with an IPS matrix up to 5.7″ diagonal are more than enough. For AMOLED screens, which, firstly, have a PenTile subpixel structure, and secondly, may support Google VR virtual reality glasses (by the way, what percentage of users actually found this useful?), the justification for QHD resolution can still be as follows: then argue.
The iPhone X stands somewhat aside with a resolution of 2436 × 1125 - however, this, in fact, is not much different from Full HD. For comparison: the screen resolution of the Samsung Galaxy S8 is 2960 × 1440, that is, one and a half times more pixels than the iPhone X.
Now imagine that we compare the performance of an iPhone 8 with its HD resolution and some Nokia 8 with QHD. Did you imagine? Nokia has to process almost four times more pixels than the iPhone, which has an impact on power consumption and performance (at least in those tests that use screen output). Now I am in no way justifying the old screens that Apple with manic persistence continues to install in devices costing under a thousand dollars, but simply focusing on the fact that the performance and energy efficiency of devices with low-resolution screens, even all other things being equal, will be higher than smartphones with QHD screens.
The manufacturers also suspected something like this. Thus, the Sony Xperia Z5 Premium, the screen of which (by the way, IPS, useless for VR purposes) has a physical resolution of 4K (in fact, no, even here the marketers deceived), but the logical one is “only” Full HD, which allowed the manufacturer and consumer cheat, and not kill performance too much. Samsung did a similar thing, allowing the use of lower logical resolution on screens with a high pixel density. Obviously, the interests of marketers run counter to the interests of both the users of these devices and the company’s own developers.
Instead of a conclusion: does our phone need 64 bits?
Are 64-bit processors really needed? mobile devices? After all, 32-bit computing cores have their advantages. Such processors can be faster than 64-bit processors due to shorter instruction lengths due to shorter address lengths, and as a result, they are less space-intensive random access memory; they can implement a shorter command queue, which can also provide performance benefits in certain scenarios.
Some of these advantages will remain theoretical, but in a number of modern use cases it is no longer possible to do without support for ARMv8 commands. This includes stream encryption, real-time HDR stitching, and many other subtle things. Be that as it may, processor manufacturers have switched to 64-bit cores with ARMv8 support, and this is a fait accompli.
But smartphone manufacturers are in no hurry to switch to 64-bit operating system builds.
Thus, in nature there is not a single smartphone running Windows 10 Mobile in which the operating system would run in 64-bit mode. Both the Lumia 950 (Snapdragon 808), the Lumia 950 XL (Snapdragon 810), and even the relatively recent Alcatel Idol 4 Pro (Snapdragon 820) run a 32-bit build of Windows 10 Mobile.
Manufacturers of Android phones are not far behind. For example, Lenovo, which produces smartphones under the Motorola brand, has only two devices with the “correct” 64-bit Android: these are the flagships of the Moto Z line ( regular version and the Force variety) and Moto Z2 Force. All other devices - the budget Moto G5 on the Snapdragon 430, and the fresh sub-flagship Moto Z2 Play on the Snapdragon 626 - operate in 32-bit mode.
A number of devices from other manufacturers (for example, BQ Aquaris X5 Plus) use the powerful Snapdragon 652 in 32-bit mode. Need I say that such devices do not squeeze the maximum out of the available hardware capabilities?
On the other hand, not everything is perfect at Apple either. Even 64-bit applications compiled into native code, due to backward compatibility requirements, are forced to be limited to the command set available in the company's earliest processors - the Apple A7 from 2013. But the ART bytecode compiler, which has been used in Android since version 5, has no such problems: application bytecode is compiled into optimized native code that uses all the instructions available on the current hardware.
However, we will live with what we have. For maximum processor core performance and guaranteed optimization, come to Apple. The same thing, only one and a half to two times worse and the same amount cheaper - to the host of manufacturers of Android handsets.
After a series of legal battles with Samsung, Apple has finally taken the long-awaited step and decided to reduce its dependence on the Korean manufacturer for chip printing. After the iFixit team took apart the new Apple smartphones, the Chipworks team did the same, trying to identify the manufacturers of the various components of the iPhone 6.
The publication itself is quite interesting and is replete with photographs of a lot of components, but the most interesting are the conclusions regarding the A8 processor. Experts are confident that Samsung is no longer the manufacturer. Chipworks employees believe that the new contract manufacturer is most likely Taiwanese TSMC, but do not undertake to confirm this 100%. However, the gate dimensions are similar to those of Qualcomm's MDM9235 chip, which is also manufactured using TSMC's 20nm process.
The A8 chip itself includes 2 billion transistors (twice as many as the A7), but its area is 89.25 mm 2 (8.5 × 10.5 mm), that is, it is 13% smaller than the A7 (102 mm 2) . Apple claims the CPU is 25% more powerful and the graphics are 50% faster. At the same time, the chip is 50 times more productive in CPU tasks than the single-chip system used in the original iPhone, and 84 times more productive in graphics tasks. Finally, Apple says the chip is 50% more energy efficient than the A7, which should benefit battery life.
The Chipworks team also confirmed that the A8 SoC is complemented by only 1GB of DRAM, while many current flagship Android devices use 3GB of RAM. Based on the code on the particular chip obtained by Chipworks, it was only packaged six weeks ago (probably in Taiwan) before passing through Foxconn's production lines in China and ending up on a US store shelf in Ottawa.
Users of the iPhone 6s note the amazing speed of the device and the high smoothness of the interface. It's not surprising that Apple equips its flagship products, be it smartphones or tablets, with powerful hardware solutions. The best of these at the moment is the A9 processor.
The A9 SoC or system-on-chip is Apple's own design. The 64-bit processor is manufactured using 14- or 16-nm technology by two contractors: Samsung and TSMC. To the credit of the company's engineers, it should be noted that synthetic tests rank the A9 first in terms of performance. But it wasn’t always like this, let’s remember how it all began.
Until 2010, Apple was forced to use Samsung's developments. But the release of such a revolutionary device as the iPad required a radically different approach. The result was the release of Apple's first proprietary mobile microprocessor integrated into a tablet computer. The A4 chip operated at a frequency of 1 GHz and had a maximum power consumption of 500-800 mW. It was based on the ARM Cortex A8 architecture and was produced using a 45 nm process technology. As it turned out later, Steve Jobs' decision to release his own hardware platform turned out to be a major strategic step.
According to analyst Stephen Cheney, despite the success and strong position of Intel in this market, Apple was able to displace the famous electronics manufacturer. Few believed in the success of Jobs' team, and Microsoft head Steve Ballmer openly laughed at the original iPhone. However, Apple management was determined to provide itself with processors.
In 2008, the apple company acquired small organization PA Semi for $278 million, known for its energy-efficient system-on-chip designs. In subsequent years, several landmark acquisitions were also made, which ultimately brought Apple to market leadership.
It is important to remember that the hardware of the iPhone and iPad is closely related to the software. The iOS mobile operating system was able not only to conquer the market in the shortest possible time, but also to provide jobs for hundreds of thousands of programmers. Even competing companies earn much more from Apple's customers than from their own. Take, for example, such popular services as YouTube, Google Search, Google Maps, which bring multimillion-dollar profits to the Internet giant.
But the advantages of the A-series chips do not end there. Optimization plays a huge role in user experience software. Competitors can buy or license technologies developed by Apple, introducing them into their devices along the way. However, the full functionality of Apple products can only be ensured by the program code that underlies iOS. It is for this reason that implementing the 3D Touch function on Android devices in the form that exists on the iPhone 6s will be extremely difficult.
Apple continues to thrive thanks to the vision of its founder and mastermind. As you can see, an ardent pedant and follower of Zen practices, Steve Jobs turned out to be a real visionary.