Is the A14 chip to be used on the iPhone 12 strong? | Hard Philosophy

It has a full-screen appearance similar to the iPad Pro, fingerprint recognition keys on the side, and 5 color combinations… It is not excessive to say that this year's iPad Air is the most changed generation in the history of the series.

Not to mention, it is also the first to carry Apple's A14 chip.

This may be more suspicious of the positioning gap between it and the iPad Pro than the similar appearance.

▲ The previous generations of A series chips are in a leading position in process technology

The A-series chips are of extraordinary significance to Apple, not only because they are Apple’s latest achievements in the field of silicon technology every year, but also because of the many new features and capabilities on the iPhone and iPad, which are also supported and driven by it.

Because of this, the appearance of successive iPhones is often the moment when a new generation of A-series chips are officially on stage.

▲ iPad Air 4 is equipped with Apple's latest A14 chip

But this year is different. Apple chose to give the starting rights of A14 to iPad instead of iPhone.

The last time Apple would do this was when the iPad 2 was released 9 years ago. At that time, it was also the first device equipped with Apple's A5 chip, nearly 6 months earlier than the iPhone 4s of the same year.

If it were not for the impact of the epidemic that forced the iPhone’s listing to be postponed for a month, it would probably be difficult for the iPad Air 4 to enjoy this "first release award."

The world's first 5nm process chip, but not the strongest performance?

Let's take a look at how Apple officially introduced A14.

First of all, it is the world's first chip designed based on the 5nm process technology, which has greatly increased its number of transistors, from 8.5 billion A13 to 11.8 billion.

▲ A14 still continues the same CPU core design as A13, including 2 large cores focusing on performance and 4 small cores focusing on energy efficiency

Specific to each module part, the A14 still continues the same 6-core CPU and 4-core GPU design as the previous generation. Compared with the A12, the performance of the two parts is increased by about 40% and 30%, respectively.

Another improvement is the A14 neural network engine: the new 16-core architecture directly doubles the number of 8 cores compared to the A13, which can provide up to 11Tops of computing power.

In addition, this time Apple also introduced a machine learning accelerator structure similar to the A13 for the A14, allowing the iPad Air 4 to obtain a faster matrix operation speed.

▲ Apple's official introduction to A14

From the data point of view, the improvement of the A14 is still very eye-catching. Considering that the iPhone 12 will also be equipped with this chip in the future, it is now equivalent to let us see the performance improvement of the iPhone 12 in advance.

But after a closer look, you will notice the difference. For example, the comparison data of A14 on CPU and GPU are based on the previous year's A12 instead of last year's A13, which is not consistent with common sense.

▲ Apple chose to compare A14 and A12, mainly to reflect the performance difference between the two generations of iPad Air

According to the news that Ai Faner got, Apple will do this because of the chips of the new and old iPad Air. After all, only the comparison of similar products can achieve absolute fairness and justice, not that Apple deliberately avoided A13.

However, the performance comparison between A14 and A13 still attracts the attention of hardware enthusiasts.

At present, from the speculation of professional media such as AnandTech , the CPU/GPU part of the A14 is only 16% and 8% higher than the A13, but compared to the A12, the A13 has an increase of about 20% in both parts.

This means that even if the A14 uses a new generation of 5nm process, the improvement is likely to be inferior to the previous generation of A13, especially the GPU graphics processor, which has an improvement of less than 10%, which is quite "conservative."

▲ This year’s iPhone 12 must not only support 5G, but also a new small-screen model, which has unbalanced power consumption and is prone to battery life problems.

What is the reason behind it, we do not yet know. It is speculated that Apple may have deliberately slowed down the performance upgrade of this generation of A14 based on the power consumption of the 5G iPhone. We can only learn about this after the iPhone 12 is launched.

It is also said that the current 5nm process yield of TSMC has not yet reached the standard, which indirectly affects the performance release of the A14. Whether it is really the case, you need to wait for further real chip disassembly and analysis.

▲ At present, the information of A14 is still very limited, we still have to wait for the follow-up chip disassembly and test data

Another problem derived from A14 is the performance comparison between iPad Air 4 and iPad Pro.

Since the current iPad Pro is still equipped with the A12Z chip, many people believe that the iPad Air 4 equipped with the latest generation of A-series chips is "following" and surpasses the old generation of iPad Pro in performance.

However, the theory of "the higher the number, the stronger the performance" does not apply to A series chips with suffixes such as X and Z.

Objectively speaking, the upgrade of the process technology and the gap in energy consumption will indeed enable the A14 to have stronger single-core performance.

​But we can't ignore the fact that A12Z itself has an 8-core CPU and an 8-core GPU. This multi-core structure, as well as the advantages of memory bandwidth, can still make the iPad Pro better than the iPad Air 4 in multi-threading and graphics processing capabilities.

▲ The A12Z on the iPad Pro uses an 8-core graphics processor to cope with heavy tasks such as 4K video editing, 3D design, and AR augmented reality

Therefore, as long as Apple has no plans to release A14X or A14Z, A12Z will still be the A-series chip with the strongest overall performance on Apple mobile devices. This is estimated to be unchanged after the release of iPhone 12.

The computing power of machine learning is greatly improved. Why should Apple pay attention to it?

In the last two generations of A-series chips, the neural network engine is a module that is often mentioned by Apple.

The A14 is no exception this time. The number of cores of the neural network engine has been directly doubled, from the 8 cores of the A13 to the 16 cores of the A14. It is estimated that 40% of the number of transistors has increased, and a considerable part of it is used here.

And the nearly twice the increase in machine learning computing power brought about by this is obviously much more dazzling than the upgrade of CPU and GPU, let alone the machine learning accelerator applied to the processor.

▲ Compared with CPU and GPU, A14's improvement in machine learning computing power will be more obvious

However, many people do not know what experience this neural network engine will affect.

It is difficult for us to attribute it to simply "improvement of mobile phone fluency", because it is not like CPU and GPU, which can have quantitative data as a basis for comparison; even if the computing power is doubled, it will not let your game, from The original 30 frames rose to 60 frames.

However, many of the features in recent generations of iOS systems are inseparable from machine learning technology.

For example, taking pictures, the DeepFusion three-in-one image processing technology introduced by the iPhone 11 Pro last year is to realize image synthesis through algorithms trained by machine learning, and finally generate a higher-quality photo.

▲ Image recognition, text overview, semantic analysis, etc. all involve the application of machine learning technology

Furthermore, the iPhone’s photo album can automatically categorize the gallery based on facial features; Apple Music can automatically recommend your favorite songs based on our listening preferences; and the iPhone’s power management system will also actively learn from us. Using habits, extending and optimizing battery life, are behind the participation of machine learning technology.

In this iPad Air 4, Apple also introduced two cases using A14 machine learning technology.

One is the djay Pro application, which allows the iPad Air to call the front camera to capture and recognize the gestures of musicians playing DJ, and realize the characteristics of "DJing in the air".

Considering that the iPad Air 4 is not equipped with a Face ID component that can perceive depth scenes, this kind of action recognition can only be achieved by machine learning solutions.

Another case is the photo editing application Pixelmator. With the help of machine learning capabilities, it can automatically repair cropped and enlarged photos.

Even if the resolution of the photo itself is very limited, machine learning can still fill in missing pixels and improve edge jaggedness and image texture.

Therefore, from the above several examples, we can see that the neural network engine of the A series chip seems to be "not strong", but in fact, it exerts a lower level of influence and is not easy to be noticed by us. Many functions are actually related to Our experience is closely related.​

Last year, former Apple executive Phil Schiller also said in an interview with Wired that there are basically no areas where machine learning is not used in the functions of the current iOS system. Whether it is the impact on battery life or performance optimization, it has been running continuously in places you don't know.

That being the case, by developing stronger computing power, these tasks can obviously be run more efficiently; and Apple can also avoid the risk of personal data collection, allowing devices to rely on local computing power instead of networking to complete more serious, More complex tasks.

A series chips are no longer exclusive to iPhone and iPad

Will A14 be Apple's only chip on the 5nm process technology? We don't think so. Even if this time Apple focuses on machine learning computing power, rather than CPU and GPU performance, we still have the opportunity to see a more powerful derivative version on the next-generation iPad Pro, or even Mac.

▲ At this year’s WWDC, Apple launched the Mac mini development machine equipped with the A12Z chip for developers

We have also mentioned in the previous article that Apple has started to maintain three "high, middle and low" product lines in the iPhone, iPad and other product lines, or even more than three, with products with different names and different price points, but not all of them. All products need the strongest chip performance.

The changes in the audience have made it difficult for Apple to cover the entire product line with one A-series chip per year. Instead, it has to make trade-offs based on the performance requirements of mobile phones, tablets and other devices.

▲ To widen the gap between Pro-level and mass-level products, the upper level of the chip needs to be differentiated

In recent years, Apple has also consciously expanded the branches of A-series chips. For example, for the A12 generation alone, we can see two versions of A12X and A12Z, the latter two have been enhanced in the number of CPU and GPU cores. , To support the performance required by the professional field.

Nowadays, even if it is the same generation of A-series chips, Apple will adopt the method of shielding the core or opening the core according to the yield status (for example, the A12X shields a GPU core, and the A12Z is ​​completely open) to open different products In the A14, a similar situation may repeat itself again.

Not to mention that in the future, when Apple’s Mac product line breaks away from the X86 architecture and is included in the jurisdiction of the A series of chips, will we see more A series with suffixes such as X, Y, Z, L, and M? Is the revised chip distributed on all Apple devices?

Perhaps the unification of software and hardware that Apple wants will eventually have to rely on A-series chips.

Source of title picture: Wccftech

#Welcome to follow Aifaner's official WeChat account: Aifaner (WeChat ID: ifanr), more exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo