In-depth analysis of Apple's A12X chip: unimaginably powerful!

Source: The content comes from "Weiphone", thank you.

The new iPad Pro released by Apple not long ago has attracted much attention. It has a subversive design, Face ID, the new generation of Apple Pencil and USB-C interface... Almost every new feature can be foreseen. And what Apple is most proud of on this device is probably the powerful heart of the iPad Pro - the A12X chip. Recently, the hardware review website arstechnica tested this most powerful chip in Apple's history, and we will talk about it with you based on the running score results.

Even if you think that iOS devices are not perfect from various critical angles, the CPU and GPU of A12X are undoubtedly in the leading position in the field of mobile devices. This is the achievement of Apple's self-developed chips.

The A12X chip is closely related to the A12 chip in the iPhone XS/XS Max and iPhone XR. Although Apple generally does not provide specific details about its chipsets, we can roughly analyze the components of this latest A-series chip:

• CPU (Central Processing Unit): Executes most instructions that are not specific to other processing units.

• GPU (Graphics Processing Unit): Processes graphics, from screen displays to 3D gaming effects and augmented reality object resources.

• Neural Engine, also known as NPU (Neural Processing Unit): handles tasks related to neural networks and machine learning.

• IMC (Integrated Memory Controller): effectively manages data moving in and out of memory.

• ISP (Image Signal Processor): Analyzes captured photos, processes and optimizes the image created in real time.

• SEP (Secure Enclave Processor): handles sensitive data (such as biometric identifiers)

• Other components: such as display engine, storage controller, HEVC decoder and encoder, etc.

The heart and soul of A12X is still the CPU, GPU and neural engine. Let’s talk about these three important parts separately.

CPU

The iPad Pro's CPU has eight cores—four focused on performance and four on efficiency. Unlike earlier Apple chips, all of the A12X's cores can be active at the same time, marking the first time an iPad has been able to drive so many cores at once.

Apple's self-designed performance controller allows the device to call all cores at the same time. When running a large number of threaded tasks, such as professional software and professional workflows, multi-core parallelism will maximize efficiency. This is an important reason why users will see a 90% performance improvement over A10X.

As for single-core performance, Apple claims that the A12X is 35% faster than the A10X.

In order to get a more intuitive result, we compared the iPad Pro with A12X and other devices. The following is the device list:

GeekBench 4 (single core)

GEEKBENCH 4 (Multi-core)

In the running score comparison with A12 (iPhone XS), A12X did not show any advantages in comprehensive, cryptographic operations, integer, and floating point, but the score difference due to cache became an important factor in promoting A12X performance.

Apple has always focused on optimizing heterogeneous computing and cleverly using specialized types of processors to match specialized tasks. So even though the A12X itself is related to ARM's big.LITTLE architecture, Apple's extensive work in optimizing heterogeneity makes the A12X perform well on multiple cores.

A multi-core comparison of the 2018 iPad Pro and recent MacBook Pro models is enough to illustrate the efficiency of its architecture. With the exception of the latest 15-inch MacBook Pro equipped with an 8th-generation Intel Core i9 CPU, the iPad Pro performs better than every MacBook Pro we have tested.

GPU

The A12X's GPU has seven cores. This is the first time that Apple's hardware department has independently designed and implemented a 7-core GPU. It improves the graphics performance of the A10X by 100%. The graphics processing capability is at the same level as the built-in GPU of the Xbox One S. Moreover, it is a completely fanless design.

The A12X’s GPU is undoubtedly the best in the mobile space, far ahead of the graphics processing capabilities of the iPhone XS/XR A12 chip. However, it has not been able to step into the desktop field like the CPU and compete with MacBooks - at least not in benchmarks.

Smartphones and tablets have always been at a disadvantage in terms of graphics performance. At its launch event, Apple said the graphics processing capabilities of the A12X were on par with those of the Xbox One S. However, this is just an entry-level game console that usually runs AAA games at 900p resolution, which is much weaker than the Xbox One X that can run 4K resolution, and is also inferior to the PS4. Comparing tablets with game consoles is not necessarily an accurate approach, but it can indirectly reflect the catch-up of smart mobile devices in terms of game graphics.

GFXBENCH 4 Graphics Score:

NPU

The Neural Engine is designed to accelerate machine learning-related tasks performed locally on the device in a more efficient and high-performance manner than on the CPU or GPU. It is an important part of the A12X and can be said to be Apple's biggest achievement in the A series of chips so far.

In fact, the neural engine in the new iPad Pro cannot be compared with the previous generation, because this is the first iPad core with a neural engine. It is currently difficult to obtain results through quantitative scoring of NPU performance. We can only comment on the iPad Pro functions it involves.

The Neural Engine is used to perform certain specific tasks, including but not limited to: recognizing faces through Face ID and the TrueDepth sensor array in iPad Pro; scanning images and powering the search feature in the Photos app; processing voice and numerous augmented reality-related tasks, etc. The experience on modern iPhones and iPads depends largely on the Neural Engine.

Apple has kept details of its architecture secret, other than the fact that the neural engine in the A12X has eight cores. It powers many of the native features built into the iPad Pro, and third-party app developers can use it in various ways through a software development API called CoreML.

Apple has also given the neural engine an important task, performing machine learning on local devices. Unlike most technology companies' cloud-based machine learning models, Apple focuses more on evolving AI locally, although this process may be much slower than extracting user data from millions of devices in use and running it on a huge cloud computing network.

In addition, the new iPad Pro also carries Apple's ambition for AR. If the ARKit platform wants to develop better, it also depends on the follow-up of the neural engine. In fact, we have seen a surprising scene at the press conference on October 30: through the processing power of A12X combined with Project Aero software, iPad Pro can present each layer created in PSD as an independent AR object on the screen, and each layer is separated into a lifelike landscape, which is quite attractive to creators.

future

With the step-by-step leapfrogging of A11, A12, and A12X, Apple has become the leader in mobile SoC. However, the future plans of Apple's self-developed chips are not so clear, and whether the future Mac series can use Apple's own chips is still a topic of discussion. But when you hold a mobile device with a 7nm process chip, a CPU comparable to most laptops, a GPU that can beat 2/5 of the game consoles on the market, and a comfortable weight and thickness, then at least a small part of the future of mobile chips has been touched.

Powerful hardware is here, but without powerful software to support it, everything is just empty talk. We don’t want to see the iPad Pro with A12X only being able to use Final Cut, Xcode and Logic for a long time after its launch. We even doubt that there are any mobile games on the market that can squeeze out the full power of A12X. Fortunately, several heavyweight tools from Adobe for iPad are on the way, but these can only fill one corner. If you want to see the real huge potential of iPad Pro, you have to create a software ecosystem that is as powerful as the hardware.

Today is the 1760th issue of content shared by "Semiconductor Industry Observer" for you, welcome to follow.

Moore Elite is a leading chip design accelerator that reconstructs semiconductor infrastructure to make it easier for China to make chips. Its main businesses include "chip design services, supply chain operation services, talent services, and enterprise services". It covers more than 1,500 chip design companies and 500,000 engineers in the semiconductor industry chain, and has precise big data on integrated circuits. It currently has 200 employees and is growing rapidly. It has branches and employees in Shanghai, Silicon Valley, Nanjing, Beijing, Shenzhen, Xi'an, Chengdu, Hefei, Guangzhou and other places.

Click to read the original article to learn more about Moore's elite