Network Security Internet Technology Development Database Servers Mobile Phone Android Software Apple Software Computer Software News IT Information

In addition to Weibo, there is also WeChat

Please pay attention

WeChat public account

Shulou

Enter Apple's most mysterious chip lab, review Apple's past CPU architecture, and show the growth path of 3 trillion companies.

2024-07-21 Update From: SLTechnology News&Howtos shulou NAV: SLTechnology News&Howtos > IT Information >

Share

Shulou(Shulou.com)12/24 Report--

Xin Zhiyuan reports

Editor: alan run

Apple created the most successful product in the history of capitalism, becoming the first company with a market capitalization of $1 trillion. Let's explore Apple's most mysterious chip research and development department and review Apple's CPU architecture, from 68k in 1984 to self-developed chips.

Recently, Apple finally unveiled the chip research and development center, the most mysterious department within the company, and the media entered this department for the first time to get a glimpse of the birthplace of all Apple's products.

Since 2006, almost all the chips that drive Apple's new business have been born here.

The latest achievement here is Apple's recently released M3 chip.

With the blessing of 3nm process, the performance of M family series chips has been greatly increased.

Apple created the most successful product in the history of capitalism, becoming the first company with a market capitalization of $1 trillion.

Through the best-selling products such as the iPod, unparalleled branding, and the unique charm of Steve Jobs, we have succeeded in making technology stylish and elegant.

Apple's obsession with hardware optimization is impressive, changing the CPU architecture three times since the release of Mac in 1984.

But it's not an easy thing.

Whenever a technology company announces a change to the CPU architecture, there are widespread questions about whether these companies can survive the abandonment of the entire software ecosystem.

At that time, the suspicion that software was still stored on floppy disks was even more serious.

John Dvorak, a famous technology columnist, once said that transforming the PowerPC architecture into Intel's x86 architecture is a prelude to the Windows world for Apple. It opens the door for Apple to communicate more deeply with Windows.

1984 Motorola 68k

The internal chip of the Motorola 68k CPU in 1981, Apple was stumbling.

At first, the Macintosh was a relatively cheap computer, and under Jobs, Macintosh began to focus on one thing: beating the LISA team.

Jobs introduced a cutting-edge graphical user interface and asked his team to use state-of-the-art hardware.

In the field of PC, the choice of CPU is crucial.

Very early personal computers used 8bit's CPU.

However, in the 1980s, if you want to design a powerful PC computer, you need to choose the modern 16bit processor architecture.

In fact, at that time, there were three main choices: Intel 8088, Zilog Z8000, or Motorola 68k.

CPU and registers let's review the concept:

A CPU is a device that moves data from computer memory (RAM) to fast temporary memory (registers), runs operations on that data, and then moves the output back to memory.

The basic architecture registers of CPU are the smallest units in electronic memory-they hold only a few bits in the core of the CPU. CPU operates on the data according to instructions (computer programs), and the operation bits are 1 and 0.

Arithmetic logic unit (ALU) these operations are performed by an arithmetic logic unit (ALU). This is basically a collection of circuits that perform simple, specialized tasks, such as:

Add the binary numbers, for example, 0010 + 0101 = 0111

Perform logical operations, such as NOT 0000 = 1111

Move bits, such as 0011, move 1 bit to the left to 0110

The control unit of the CPU decodes one instruction at a time to determine which data should be moved to which register and which register data should pass through which ALU circuit.

Perform these operations many times very quickly, which accumulates the output, such as matrix multiplication, collision physics in video games, or rasterizing image data into screen pixels.

Let's go back to Apple.

Intel vs Zilog vs Motorola assumes that you are the vice president of hardware at Apple and are giving a presentation to Jobs.

First of all, the Intel 8088 has a 16-bit microprocessor and supports 640kB RAM,IBM PC with this chip architecture, so it has a strong existing software ecosystem, and the low-end price is about $35 (in 1983 dollars) because of Intel's huge economies of scale.

While the Zilog Z8000 is a pure 16-bit microprocessor that supports 8MB RAM, few large competitors use this architecture, the smallest software ecosystem, and the mid-range price point is about $55.

Finally, the Motorola 68k has a 16amp 32-bit microprocessor that supports 16MB RAM,Atari and Commodore using this chip architecture, as well as some existing development ecosystems. But previously established supplier relationships with Motorola through Apple I, Apple II, and LISA.

Which chip architecture would you choose?

Overall, the choice of Motorola 68k processors seems forward-looking, which explains why the 1984 Macintosh is different from other computers.

Although it lags behind the mainstream IBM PC in terms of ecosystem development and compatibility, this choice brings differentiation to the brand.

More importantly, 68k has an almost completely orthogonal instruction set, which means that almost every CPU operation can be performed on almost every register, while many competing CPU limits the register selection of instructions.

This orthogonality makes CPU easier to program and is an ideal choice for fostering a burgeoning software ecosystem.

In the end, the memory addressing range of 16MB becomes extremely important, and Macintosh reserves the memory of the top 12MB for use by the operating system, while sharing 4MB memory with different software applications.

If you look at the storage space available in 16GB iPod Touch in 2012, you will find that there is very little free space.

1994 PowerPC

The internal chip time of the PowerPC CPU is 1994. Steve Jobs (Steve Jobs) was ousted by Apple eight years ago and is now busy inventing Pixar and NeXT.

Apple is losing relevance.

In 1994, when Steve Jobs had left Apple to start a business with Pixar and NeXT, Apple gradually lost its market influence.

Compared with their previous competitor IBM, they are experiencing Microsoft's competition, which is a long and painful process.

During this period, Intel and Bill Gates (jokingly known as the "devil" in the 1990s) formed an evil alliance called Wintel, which gave the two companies a near monopoly on the market.

This is not to underestimate the power of the x86 chip architecture. Intel has gained market dominance with its 100MHz speed and excellent energy efficiency performance.

However, Motorola 68k series chips, which led Macintosh into the 1990s, failed to keep up with the development.

Faced with the threat of monopoly, Apple has partnered with longtime partner Motorola and unlikely ally IBM to use unity against capitalist forces.

As a result, the AIM (Apple, IBM, Motorola) alliance came into being. They realized that there was an important weakness in the x86 architecture: it adopted the CISC architecture.

In order to deal with this problem, AIM adopts RISC strategy, namely PowerPC.

Apple has developed an emulator that enables PowerPC to emulate Motorola CPU. This means that it can convert instructions from one architecture to another in real time.

Obviously, this will incur performance costs. Fortunately, however, because PowerPC CPU is so powerful, this kind of emulation is usually not a big problem for users who upgrade hardware.

Another strategy adopted by Apple is to use wide binaries during the transition period. In this way, the software can contain code compiled for 68k and PowerPC architectures.

As a result, engineers can publish a single application that contains two separate binaries that can run on both Mac CPU platforms.

At that time, the capacity of hard drives was usually only 80MB, which became very annoying. As a result, binary split tools have emerged so that end users only need to save files that apply to their devices.

Overall, Apple's transition was successful. The transition from 68k to PowerPC has brought a huge performance improvement. Simulation and wide binaries make the software ecosystem transition smoothly.

Unfortunately, however, the Wintel Alliance has hardly been affected. With the release of Pentium and Windows 95, their market dominance has reached an unprecedented level. Windows is becoming the default computing platform, turning the ICT curriculum in schools around the world into how to use Microsoft Office.

Now that they have a solid hardware platform, Apple's outdated System 7 Mac operating system has become a major obstacle.

They tried an internal project to create a competitor to Windows, but failed, so they had to solve the problem through acquisitions, that is, simply buying a new operating system.

This laid the foundation for Apple's acquisition of NeXT and the return of Steve Jobs.

In 2006, Intel x862000s, Apple regained its magic again, and after the return of King Jobs, the landmark Mac OS X and the success of iPod transformed the beleaguered Apple into a consumer electronics giant.

However, the PowerPC architecture at this time has failed to keep up with the pace of the Intel x86 behemoth.

The market for laptops is growing, and the power consumption and heating of the PowerPC CPU are at a disadvantage to build the ultra-thin laptops that Jobs envisioned.

At the 2005 Global developer Conference, Jobs officially announced the transition from PowerPC architecture to Intel. By 2006, it is expected that the performance per watt of high-end Intel x86 processors will be nearly five times that of PowerPC, and the clock speed will be nearly 1.5 times faster.

So what makes the Intel x86 architecture so good? Here we briefly discuss some concepts in CPU design and the innovations made by Intel in all aspects of CPU at that time.

CPU cache (Caches):

After CPU gets the data from the memory (RAM), it needs to put it into a register on the chip to perform the operation. However, the clock frequency of CPU is much higher than that of RAM, so RAM has become the bottleneck of system speed.

Therefore, CPU has developed a cache on chip (Caches) to store a medium amount of data. It acts as a buffer zone closer to the chip itself and allows faster access to the necessary data.

The caches themselves are hierarchical:

The storage layering in the computer system follows the above pyramid structure. From bottom to top, the capacity of the storage medium is getting smaller and smaller, but the speed is getting faster and faster, which makes full use of the "locality" principle of the computer program and greatly improves the overall efficiency.

Branch Prediction (Branch Prediction):

Branch instructions are used to process conditional statements. Smart CPU uses statistics to guess where the code is going and tries to keep the instruction pipeline filled for maximum utilization.

The mechanism involves hardware algorithms directly built into the CPU circuit. A buffer called a branch history table caches the most recent branch result.

Superscalar Architecture (Superscalar Architecture):

Superscalar architecture is the ultimate goal of multitasking. Superscalar CPU can execute multiple instructions simultaneously in a single clock cycle.

This architecture works because arithmetic, moving memory between registers, and floating-point operations require different circuits on the ALU.

To embrace the major transformation of Intel x86, Apple has once again adopted its time-honored transition technology to achieve a smooth migration of CPU architecture.

Apple introduces common binaries built for both CPU architectures, which can be set up through a simple Xcode build configuration.

With the help of hardware and software, Apple delivered the perfect work, realized Jobs' dream of being "ultra-thin" and brought Apple into a new era.

In 2020, Apple Chip knows Jobs and people who know Apple will understand their ultimate competitive advantage: the tight integration of hardware and software.

The iPhone released in 2007 uses the ARM CPU provided by Samsung. Starting with iPhone 4, Apple began to design its own chip A4.

In 2008, Apple bought PA Semiconductor, a CPU design company known for its high-end, low-power processors, for $278 million.

Ten years later, in 2018, Apple bought part of European chip designer Dialog for $300 million.

Finally, in 2020, Apple announced the third great transformation of the Mac CPU architecture-- M1-- which heralded the arrival of the Apple Silicon era.

So, compared to the Intel X86 architecture mentioned above, what is M1's secret weapon?

First of all, the M1 chip uses a heterogeneous computing strategy, which means that dedicated components are used for specific workloads. The components of M1 SoC are dedicated to many computing tasks.

Secondly, M1 chip has a unified memory architecture shared between GPU and CPU. Apple allocates the same memory blocks (RAM and L3 cache) to be shared between the two processors, and its format can meet both the high throughput required by CPU and the large blocks of data preferred by GPU.

Traditionally, when sending data to an external GPU for processing, the CPU usually needs to copy the data into GPU-owned memory before it can be picked up for processing, and the unified memory architecture solves this problem.

We discussed the superscalar architecture earlier, which enables the CPU kernel to read, decode, and dispatch multiple instructions at the same time. The M1 chip, with its RISC architecture, enables Apple to take it to a new level through out-of-sequence execution.

The basic M1 chip has eight decoders, and the CPU kernel fills these decoders at the same time in each clock cycle. These instructions are distributed to various dedicated circuits in parallel.

Finally, all the components on the M1 chip are physically very close. Even if the electrical signal moves at a lightning speed, the operation will be faster when the transmission distance is shorter.

With the blessing of the above designs, Apple's own chip performance continues to improve and eventually surpasses it.

Perhaps the reason why Apple is Apple is that it has always been the ultimate control over hardware and software, so what will happen to Apple's future in today's open source era?

Reference:

This article comes from the official account of Wechat: Xin Zhiyuan (ID:AI_era)

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.

Views: 0

*The comments in the above article only represent the author's personal views and do not represent the views and positions of this website. If you have more insights, please feel free to contribute and share.

Share To

IT Information

Wechat

© 2024 shulou.com SLNews company. All rights reserved.

12
Report