Shulou(Shulou.com)11/24 Report--

This article comes from the official account of Wechat: developing Internal skills practice (ID:kfngxl). Author: Zhang Yanfei allen

Hello, everyone. I'm Brother Fei!

On October 16th, Intel officially released the 14th generation of Core processors. But there are still many students who do not understand the various CPU parameters released at this kind of conference. I would like to take this opportunity to give you an in-depth knowledge of CPU's model rules, intergenerational architecture, and microarchitecture.

CPU is the most important thing in the whole computer hardware and technology system. But it is a pity that more than 80% of the development students do not know much about CPU. So in the previous article, I took you through a lot of articles about managing and scheduling CPU resources in the kernel. Today, I start with the technical details related to CPU hardware to lead you to a better understanding of CPU.

In this article, I give an example of a laptop with a previously used ThinkPad x270 on hand. The CPU model on this computer is Intel (R) Core (TM) i5-7200U CPU @ 2.50GHz 2.71GHz. So what technical information is implied in this model string? Today, let's take this model as a starting point to gain an in-depth understanding of the technical points related to CPU hardware in personal PC computers.

First, Intel's CPU model rules no matter which CPU manufacturer is, in order to better manage its many models of products, but also to better let consumers understand their own products quickly. Will define a set of product rules. For Intel, its command rules are roughly as follows:

Intel divides the number of CPU into five parts: brand logo, brand modifier, intergenerational number, SKU, and product line suffix. Next, let's expand on these five parts to understand.

The first part is the brand logo. The most common x86 CPU brands include Intel and AMD. And Intel faces different market needs, including Ling Motion, Cy Yang, Pentium, Core, Zhiqiang and other sub-brands.

CORE, CELERON and PENTIUM are all brands used on home PC computers. Among them, the two brands Pentium and Caiyang have been counted as history. Pentium was launched in 1992 and has been the mainstream of PC home computers for a long time. Caiyang is a product launched in 1998 to compete with AMD in the low-cost market, which can be understood as a shrinking version of Pentium. Core was launched in 2006 and is still the mainstream brand in PC computers.

Xeon is an enterprise-class CPU, mostly used in servers and workstations. The online servers we use at work are basically products of this sub-brand.

Atom, which focuses on power saving and low power consumption, is mainly used in devices such as mobile phones and tablets. Yes, Intel has also done mobile phone CPU, but it was defeated by various processor manufacturers of arm architecture in the competition because of power consumption, and gradually withdrew from the market after 2016.

The second part is the brand modifier. The purpose of this part is to distinguish the market positioning of processors. Generally speaking, the performance of i3i5i7i9 increases in turn in CPU produced in the same era. It represents the positioning of low-end, middle-end and high-end respectively.

The third part is the intergenerational number of CPU. This part may be one digit or two digits. On the CPU we exemplified, the intergenerational number is "7". This intergenerational number corresponds to Intel's internal code-named "Kaby Lake" CPU architecture launched in 2016. In general, the higher the CPU intergenerational number, the newer the point in time when the architecture is launched. The performance of a single core will also be better. We will introduce it separately after "Kaby Lake".

The fourth part is the SKU number. This is the number made by CPU manufacturers to facilitate inventory management of all products. This is similar to the bar code of goods in the supermarket and the ISBN number of books, which makes it easy to query and manage the inventory of products. In general, the larger the value, the better the performance, but not absolutely.

The fifth part is the product line suffix. In laptops, H represents high performance, U series represents low power consumption, and Y series represents very low power consumption. In desktops, X represents the highest performance, K represents high performance, and T represents power optimization.

For model rules, you can also refer to Intel's official website introduction (https://www.intel.cn/ content / www / cn / zh / processors / processor-numbers.html).

Now let's go back to the CPU,Intel (R) Core (TM) i5-7200U CPU we mentioned at the beginning. The implication is that this is an Intel Cooley brand of CPU,i7, which represents the mid-to-high end, intergenerational number 7, and this intergenerational processor architecture is produced by the Kaby Lake architecture released in 2016. The SKU number is 200. The U of the product line suffix represents low power consumption, more power saving, and also means that the performance is not the strongest.

Looking at the latest release of Core CPU models, Intel (R) Core (TM) i9-14900K, Intel (R) Core (TM) i7-14700K, Intel (R) Core (TM) i5-14600K and other CPU. It also shows that it still belongs to the products under the brand name of Intel Core. I9, i7 and i5 represent the level of each CPU, respectively. 14 represents intergenerational. 900,700,600 belong to SKU numbers.

Second, the intergenerational CPU in front of the Kaby Lake, I am sure that many friends will say that they do not understand what this means. So here is a special section to explain. We mentioned earlier that each CPU model contains an intergenerational number. For example, Generation 7 and Generation 8 have just been updated to Generation 14 on October 16, 2023.

This intergenerational CPU represents different years of production, different production processes, and different architecture designs. In addition to the number of this number, there will be an English code name for each generation. For example, the intergenerational number of the seventh generation is Kaby Lake. The latest generation 14 intergenerational number is Raptor Lake Refresh. The following table shows the intergenerational information I have collected (you need to drag the table to the right on your phone to see it all).

Time Algebra CPU Intergenerational process Microarchitecture 2013 4th Generation 14nmSkylake2016 14nmHaswell 2014 Generation 6 Skylake (client) 14nmSkylake2016 Generation 7 Kaby Lake14nmSkylake2017 Generation 8 generation Coffee Lake14nmSkylake2018 Generation 9 Coffee Lake Refresh14nmSkylake2019 Generation 10 generation Ice Lake (client) 10nmSunny Cove2020 Generation 11 Tiger Lake10nmWillow Cove2021 Generation 12 Alder Lake S7nm small Core Gracemont2022 Generation 13 Raptor Lake7nm small Core Gracemont2023 Generation 14 Raptor Lake Refresh7nm large Core Raptor Cove small Core Gracemont in the above table I listed the release time, algebra, intergenerational numbering, manufacturing process, and finally micro-architecture of each generation of CPU. Because the CPU in my old computer is Kaby Lake, so I will find out the details of Kaby Lake for you.

Our CPU is dual-core four-thread. So the die diagram of the corresponding CPU (the internal physical structure of the CPU chip) is as follows

In the Die diagram, the top Memory Controller is the memory controller. What kind of memory CPU can support and how much memory it can support is determined by the memory controller in CPU. Compared with the previous generation Skylake, the memory controller of Kaby Lake increases the supported memory frequency from DDR4-2133 to DDR4-2400.

In the middle below are two physical cores, which are the most important parts of the CPU. There are also two L3 Cache. Note that although there are two L3 caches in hardware, no matter how many L3 caches are, all L3 caches are shared by all cores.

The rightmost Gen9.5 below the CPU is the integrated video card in the CPU. Since we are not going to talk about graphics cards in depth, all we need to know here is that the CPU of PC computers contains an integrated video card. Students who are interested in this graphics card can move to https://en.wikichip.org/ wiki / intel / microarchitectures / gen9.

On the far left below the CPU is the System Agent module. The module mainly includes peripheral components fast interconnection (Peripheral Component Interconnect Express) channel, image processing unit (Image Processing Unit,IPU) and so on.

Contains 20 PCIe channels in Kaby Lake. Some of these channels are used to communicate with graphics cards, and some are used to communicate with hard drives, network cards and other devices through the South Bridge.

The South Bridge chip is used to help CPU interact with peripherals. IO devices such as solid state drives, mechanical hard drives and network cards are generally connected to the CPU through the South Bridge. South Bridge will collect the data on these devices and transmit them to CPU for processing. In addition, some high-performance solid state drives also communicate with CPU directly through the PCIe channel, so the performance of such hard drives is better.

In addition, the function of IPU is to support video recording and imaging, face detection and recognition and other image related processing hardware support.

These are the main physical structures of CPU under the Kaby Lake architecture.

You can also see from the table that Kaby Lake was released in 2016 and the process is 14 nm. The manufacturing process of CPU has been improving, from 22 nm in 2013 to 7 nm today. The main benefit of the progress in process technology is the improvement of energy efficiency, the number of transistors per unit area has increased, but the energy consumption required has become lower.

As for the changes of memory controllers and PCIe channels in other intergenerational architectures, we will write a separate article later.

Third, physical nuclear micro-architecture design. In addition, each generation of CPU is very different in that the micro-architecture is different. Microarchitecture refers to the technical scheme used in the design of each physical core. Different microarchitectures have different designs for the core, such as TLB, L1, L2 and other caches, as well as the computing units within the CPU core.

From the table in the previous section, you can see that I have this Intel (R) Core (TM) i5-7200U CPU intergenerational number Kaby Lake, and the micro-architecture design is still using the core design in Skylake. So how is the physical core in Skylake designed? I went from the official website to its internal physical structure.

The above picture can be roughly divided into three large areas according to different colors.

The first area is the front end of the yellow section, also known as Front End.

The function of the front-end part is to get the instruction from memory and decode it, then put the decoded instruction into the queue and wait for the CPU back-end to process. It also includes the implementation BPU of branch prediction in CPU. It also contains L1 instruction cache and instruction TLB that parse instructions are needed.

As you can see from the above figure, the size of the L1 instruction cache in the Skylake core is 32 KiB.

The second area is the green back end, also known as Execution Engine.

The function of the back-end module is to get the front-end decoded instructions from the queue and start running. In the backend, it includes 8 ports such as Port0, Port1,..., Port7 and so on. Each port supports different micromanipulation (uOP) processing.

Port0, Port1, Port5 and Port6 support the addition of integers and floating-point numbers, Port2 and Port3 are used for address generation and loading, and Port4 is used for storage operations. Every one of them.

Because there are eight Port, the back end can handle up to eight microoperations in one clock cycle.

The third area is the purple storage system, also known as Momory SubSystem.

The L1-level Data cache is included here, and the figure also shows that its size is 32KiB. It also includes the L2 cache, which is 256 KiB in size. There are also caches such as Data TLB. These caches are located inside the CPU core. Accessing them is faster than accessing memory.

The micro-architecture design of different generations of CPU is also different, and the single-core performance will also change, which we will talk about in a later article.

To sum up today, with the help of an old CPU Intel (R) Core (TM) i5-7200U CPU @ 2.50GHz 2.71GHz. We took an in-depth look at Intel's CPU naming convention. Knowing the naming rules can help you quickly judge the approximate performance of a CPU. However, if you are an enthusiast, it is recommended that you search the ladder chart to find out the ranking of each CPU.

In addition, we also found the hardware anatomical diagram of this CPU-Die diagram. From the Die diagram, we can directly see the physical structure of the CPU. Integrated graphics cards, physical cores, and the location of the L3 cache are included in the i5-7200U. In addition, the memory controller is also integrated in CPU, through which CPU communicates with memory. There is also the PCIe channel, which is used by CPU to communicate with hard drives, network cards, graphics cards and other devices.

Finally we went deep into the inside of the physical core of Skylake. Learned about the three major components of the physical core: the front end, the back end, and the enclosure. The front end is responsible for the parsing of instructions, and the back end is responsible for the running and processing of microoperations after instruction parsing. The storage system provides L1, L2 and other caches for the front-end and back-end to accelerate memory access.

Seeing here, I believe you will have a deeper understanding of CPU. Later, we will have another article to introduce where progress has been made in different generations. Don't hesitate to share this article with your best friends.

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.