The central processor of the computer (CPU)

Центральный процессор компьютера (ЦПУ)

Today we are considering the central processor of a computer CPU (Central Processing Unit - central processing unit or CPU). This is the heart of the system unit or, if you will, its brain! In the jargon of computer programmers it is sometimes called "stone" (crystalline silicon really looks like a stone).

This is the main handler of the information arriving at the computer. The central processor performs all the necessary mathematical operations with incoming data, produces various samples from databases, archives and unzips our files, handles video encoding, processes the model of physical interaction of particles in your favorite computer game, finally!

The central processor of a modern computer does a lot, it will be easier to list what it does not do :)

Here are some photos of the CPU:

The first photo, this is a dual-core central processor of the company "Intel", the second - its back. This side is inserted into the processor socket (socket socket) on the motherboard . In these photos we see the CPU of the form factor "LGA-775". The abbreviation "LGA" is an abbreviation of English. "Land Grid Array" - the type of housing with a matrix of contact pads. Obsolete models were supplied in the "PGA" (Pin Grid Array) housings, just such an out-of-date processor is presented in the last photo above.

The introduction of the new form factor was due to the fact that the number of "legs" (pins) of the CPU of previous generations increased to such an extent that between them began to appear parasitic electrical interference, affecting the operation of the final device. The peculiarity of "LGA" is that the contacts themselves are transferred from the processor case to the socket (socket) surface, which is located on the motherboard. On the chip substrate, only the contact surfaces remain (the so-called "patches").

In older computer models, the processor installation was associated with a certain risk of bending or (God forbid) to break one of several hundred PGA legs. A terrible dream of the assembler of computers! :) Now everything is much easier.

What we see in the photos above is the outer shell of the computer's CPU. Its function is to protect the core (the silicon crystal itself) from mechanical impact, to provide the area of ​​contact with the cooling system (radiator), and also to provide an electrical contact for powering the device (photos above under number "1" and "2").

The central processor of the computer consists of a square plate of textolite, into which its core (a silicon crystal) is tightly mounted, as well as the leads of electrical contacts, plus a protective cover from above. What is under this lid we considered here.

The process of manufacturing finished chips can be described approximately as follows: layers of conductors, semiconductors and insulators are alternately applied to a thin silicon substrate (substrate) through special "masks" with slits by lithography. Sometimes the process of etching elements on a chip (through the same holes in the "mask") is used. After the end of the procedure, the substrate is sawn into squares, which are clad in a protective and heat-conducting shell, are provided with contact pads and the product is ready!

Now the market of desktop processors is divided among themselves almost only two large companies: "Intel" and "AMD". According to 2011 data, the first "held" more than 80% of this market, and the second - just over 10%. Quite another matter is the rapidly developing mobile processor market. Here there is simply a huge number of companies that issue their solutions (well, not quite "their own", but we'll talk about this in another article).

Schematically, the internal device of the CPU can be represented as follows:

Here is a visual picture of the CPU chip in the section:

And this is a powerful dual-core Athlon:

Yes, here is another photo, for completeness of the picture so to speak:

This is also a computer processor, just in a different design. There were, in due time, similar samples, a textolite board which was inserted vertically into a special connector on the motherboard. It was called (Slot A), hence the term "slot processors". Most of all, the design resembles a cartridge of a game console with a fan at the side :)

If you touch on such an important aspect as CPU performance, then it directly depends on several components and the following is composed of them:

  • its clock frequency
  • number of cores
  • number and speed of memory caches

We will analyze each of the points in more detail. The processor clock speed is measured in Hertz (Hz).

Note: Hertz (Hz) is a unit for measuring the frequency of periodic processes (in this case oscillations). For example, 1 Hertz - one such oscillation (tact) per second.

To measure the clock speed (performance) of the central processor in Hertz is inconvenient (too large numbers are obtained). Therefore, such quantities as megahertz and gigahertz are used here. Megahertz (Mhz) is one million Hertz (1,000,000 Hz). The GHz (Ghz) ​​is 1000 megahertz (Mhz) or - one billion Hertz (1,000,000,000 Hz).

According to the above, it turns out that the CPU with a clock speed of 3 Gigahertz is 3000 Megahertz or three billion Hertz! Conditionally it is possible to say so, the higher the frequency, the more instructions can be processed per unit time. According to the example described, a processor of 3 Ghz (GHz) can perform three billion operations per second.

For better assimilation - see a small thematic video:

You can see the value of the clock frequency by right-clicking on the "My Computer" icon on the desktop and selecting "Properties" from the drop-down menu. The screenshot below shows images with this information for the operating systems "Windows 7" and "Windows XP".

Also, this indicator can be seen during the initial boot of the operating system, going into the BIOS or using one of the specialized utilities, like "CpuZ". This wonderful program will show not only the value of the clock frequency, but also many other useful information.

Note: we considered the work of this program in sufficient detail here, so we will not repeat it.

Remember the famous "law" of Gordon Moore, deduced by him back in 1975: "The performance of modern processors should be doubled every 24 months!" We must pay tribute to this prediction: it was so, until some time. Manufacturers of processors simply regularly increased the clock speed of their devices (amid other improvements, in the form of parallel processing commands, expanding the list of supported instructions, reducing the process technology, etc.), which allowed maintaining the survivability of this statement.

It is clear that this could not continue indefinitely: large frequencies require radical processing of the cooling system of a rapidly heating chip. The author of the statement in 2007 said that, apparently, the "law" will not last long. The fact is that when a certain frequency threshold is reached (in the range from 4000 to 5000 Megahertz), any processors start to work unstably and require a complicated cooling system.

Overclockers ("overclockers" of central processors) with experience assert that the approximate limit of an overclocked processor with air cooling is 4000-4500 Mhz. It should be understood that these are the best samples of chips, the most successful ones from the batch, and there may be one for several dozen, plus a top motherboard that allows you to apply increased voltage to it and increase the FSB frequency, expensive (overclocking) memory with additional cooling and etc. If you install a water cooling system on the same CPU, you can raise the frequency to 5000, but not the fact that it will be possible to achieve stable operation of the device in all applications.

Note: FSB (Front Side Bus) is a high-speed interface for interfacing between the computer processor and other peripherals and modules located on the motherboard. The frequency of the system bus is the speed at which the processor core exchanges data with RAM, a discrete graphics card, hard disk controllers, etc.

Real "maniacs" of their work do not stop and this and the course is "heavy artillery" like cooling using freon, liquid metal, helium and even liquid nitrogen! The latter option allows you to "squeeze" from the unfortunate device a record 6000 Megahertz and even more! On the other hand, is it unlikely you want to work on a computer covered with ice crust? :)

Now the moment has come when the frequency and overall speed of modern computers are sufficient to solve most of the tasks of the ordinary PC user (now we omit games and serious applications for modeling something). That is why a simple increase in this indicator will no longer give such a tangible increase in speed in everyday (office) tasks, as before. Now the performance of modern PCs is largely determined by other parameters and their combination.

One of these parameters is the unification under a single heat dissipating cover of a large number of nuclei (at the moment their number can reach twelve pieces). Here arithmetic is simple: the more cores, the higher the productivity (other things being equal). After all, all processes, in this case, begin to be performed in parallel (on each of the cores), which (in theory) should significantly increase the overall speed. In practice it turns out ... differently :)

Some of the applications simply "do not know" that you can work with several cores, some do it badly and only a specially "sharpened" for multi-core applications there is a significant increase. There are applications that practically can not be paralleled. For example, office applications ("Microsoft Word" or "Open Office"). Other tasks, such as video / audio coding, compilation of program code, rendering of a three-dimensional scene, on the contrary are very sensitive to multithreaded processing and the maximum gain is obtained with this approach.

The locomotive of multi-core is considered to be the server variants of the central processors. This is "Intel Xeon" and "AMD Opteron" respectively. Server solutions are characterized by increased speed (due to a large cache) and scalability (can have several physical processors with a large number of cores inside each). Such systems enthusiasts sometimes install and at home on ordinary motherboards, but this is more for the sake of sporting interest :) Basically, similar processors are used in river servers that are mounted in special racks.

Note: (Rack - rack / shelf) RackMount - the principle of the organization of switching equipment.

So this kind of server might look different:

And here so - in the river 19-inch rack (it is also called the telecommunication rack):

There are even lockable whole telecommunication cabinets (Protective Cabinet). They can look, for example, like this:

In detail about how such servers are arranged inside, what processors they have and how the server room is organized in our work, we considered in one of the lessons.

On the basis of such decisions, so-called supercomputers are being built. For example, Intel has already released 16-core Xeons and is considering solutions with 22-24 and 28 cores. Do you understand where all this business is moving, right? So the joke of the KVN team "Ural pelmeni" about the 48th nuclear processor, pronounced in 2012, does not look like a joke any more! :)

I'm sure, over time, most applications will work effectively on multi-core systems, now, with this, not everything is so rosy. But the manufacturers of CPUs are steadily increasing this figure and now there are desktop systems with 12 cores. What for? Well, one must somehow explain to the buyer why he simply has to buy this new processor ?! :)

The third most important component of the computer's CPU is its cache. A cache is a small amount of very fast memory that is located in the core itself and serves to store intermediate computational results, and can also store copies of the most frequently used data from the computer's RAM . The cache can act as a kind of "bridge" with a speedy movement between the RAM and the CPU of the computer.

The cache is divided into the instruction cache (to speed up the loading of the machine code) and the data cache serving user requests. The latter often has several levels (Level 1, Level 2 and Level 3). Each subsequent level is larger (by the amount of memory) of the previous one, but slower in speed. Why so? I think, to reduce the cost of the final product :) But its such a design gives, - a significant reduction in the latency of CPU access to RAM. This is a kind of buffer between it and the CPU.

There are specific tasks, where the cache of the processor plays out not the last role. It is believed that this includes the process of archiving information arrays and devices with a large and fast cache to cope with it better.

As we can see, neither frequency, nor multi-core, nor big cache guarantees speeding up our work in all tasks! Somewhere it will be enough just for big speed (frequency), somewhere it will require multitasking - the operation is performed in parallel on several cores. This requires a comprehensive approach and a delicate balance between all components.

Go ahead! Since the CPU is running, an electric current is applied to it. This leads to the fact that it is heated. To avoid such an unpleasant phenomenon as overheating of the computer's processor , various cooling systems (noiseless water or based on air cooling equipped with fans) are installed on it.

Despite the constant reduction of the technological process and the optimization of power consumption, the top processor models are stubbornly storming the 200 Watt TDP bar, and some (AMD) have successfully conquered it! Can this "achievement" be unequivocally called a victory? I do not think :)

Each of the manufacturers gives their new product a code name, which characterizes a whole line or family of products based on one microarchitecture. In the recent past, such sonorous names as "Coppermine", "Wolfdale", "Barton", "Nahalem", "Prescott", "Conroe", "Sandy Bridge" were used.

It is the microarchitecture of the core that determines which of the new technologies will be incorporated into the future processor. For example: hardware (at the hardware level) support for virtualization technology (Visualization Technology), buffer overflow protection (Intel Execute Disable Bit), AMD Turbo Core, automatic controlled overclocking of the processor (analogue TurboBoost from Intel), various options for SSE instructions and 3D Now, etc.

Now it's fashionable to talk not about CPU, but about APU (Accelerated Processor Unit - accelerated processor). What it is? This is a combination on a single chip or simply under one heat distribution cover of the CPU and video card itself . Such solutions are sometimes called hybrid processors. The result is a reduction in overall energy consumption and system costs by reducing the number of components (an external graphics card is no longer needed).

It is clear that such a system can not compete with a full-fledged gaming computer, but for most tasks it is even suitable. Considering the fact that in 2006 a well-known company AMD purchased a no less famous company for the production of graphic accelerators ATI, it is logical that its APUs look more preferable (due to the graphic component). The company "Intel" has never seriously dealt with graphics, its hobby is central processors and on this field it has no equal!

What else can you say about processors in the application plan? You, as a potential buyer, will not be unpleasant to know that they can be purchased in two different variants of delivery: "Box" (Box) and "Tray" (Tray). Boxing is boxed delivery:

Let's see what's inside the box?

We see here a cooling system packed in protective plastic (from below) and the CPU of the computer itself (circled in red). Please note that in the boxed delivery, a thermal interface has already been applied to the lower surface of the cooling radiator (a heat-conducting substance in the form of three gray strips). The thermal interface serves to better transfer heat from the core of the crystal to the radiator. We can only open the package and install the design on the board.

If you decide to purchase a processor in the supply of the TRAY, be prepared for the fact that it can be carried to you in a plastic bag :) You buy separately only the chip itself, without a cooling system. What for it can be necessary? For example, I did this when I was building my home computer. Standard (boxed) cooling I did not like and I decided to install a tower system instead. Why overpay for an unnecessary piece of aluminum with a fan, which then will lie idle?

Finally a small reminder from personal experience: in modern games the processor is not the main thing. Основная нагрузка ложиться на внешнюю видеокарту, поэтому если Вы собираетесь производить модернизацию (апгрейд) своего компьютера именно с этой целью, то в первую очередь обратите внимание именно на его графическую подсистему. Почему я так уверенно об этом заявляю? Потому что сделав именно так (оставив старый процессор и купив новый GPU), я получил абсолютно нормальное быстродействие во всех играх 2015-го года!

Да, чуть не забыл! Я же хотел поделиться с Вами замечательной программой для тестирования процессоров! Она позволяет по максимуму нагрузить ЦП и выявить возможные проблемы в его работе. Нагружается не только сам процессор, но и фазы питания на материнской плате, поэтому утилита полезна вдвойне. Также она пригодится тем, кто занимается ремонтом компьютеров и, по долгу службы, вынужден устраивать своим «пациентам» стресс-тест на стабильность их работы.

Программа имеет различные режимы тестирования, а результаты ее работы Вы можете видеть в режиме реального времени в форме удобных, наглядных графиков.

Или — в виде таблицы:

Well, while this is all, we figured out with the processor, you will now know what the processor is for in the computer, and by what scheme and principle it works. If you are interested, you can watch a video about how processors are made. You like it, I advise you to look!