What the difference between the CPU from the GPU. Central and Graphics Processor Graphics Card Card

The source material for rendering is a plurality of triangles of various sizes, of which all the objects of the virtual world are developing: landscape, game characters, monsters, weapons, etc. However, the models, created from triangles, look like wire frames. Therefore, textures are superimposed on them - color two-dimensional "wallpapers". Both textures and models are placed in the memory of the graphics card, and then, when creating each frame of the game action, a rendering cycle is performed consisting of several stages.

1. The game program sends the graphic processor information describing the game scene: the composition of the objects present, their color, position relative to the observation point, lighting and visibility. Additional data, characterizing the scene and allowing the video card to increase the realistic image of the resulting image, adding fog, blur, glare, etc.

2. The graphics processor has three-dimensional models in the frame, determines which triangles included in them are in vain and cuts off hidden by other objects or, for example, shadows.

Then the light sources are created and their effect on the color of the illuminated objects is determined. This rendering stage is called "Transformation and Lighting" (T & L - TRANSFORMATION & LIGHTING).

3. On visible triangles, textures apply using various filtering technologies. Bilinear filtering provides for the overlap on a triangle of two versions of texture with different resolution. The result of its use is well distinguishable boundaries between the areas of clear and blurred textures arising from three-dimensional surfaces perpendicular to the viewing direction. Trilinear filtering using three variants of one texture allows you to create softer transitions.

However, as a result of the use of both technologies, only those textures that are perpendicular to the axis of view look really clearly. When looking at an angle, they are very blurred. In order to prevent this, anisotropic filtering is used.

This method of filtering textures is set in the video driver settings or directly in the computer game. In addition, the anisotropic filtering force can be changed: 2x, 4x, 8x or 16x - the larger the "Iks", the more clear images on the inclined surfaces. But with an increase in filtration force, the load on the video card increases, which can lead to a decrease in the speed of operation and to a decrease in the number of frames generated per unit of time.

At the texturing stage, various additional effects. For example, the overlaying of the Environment (Enironmental Mapping) allows you to create surfaces in which the game scene will be reflected: mirrors, shiny metal objects, etc. Another impressive effect is obtained using irregularities (Bump Mapping), thanks to which the light falling to the surface at an angle creates the appearance of the relief.
Texturing is the last stage of rendering, after which the picture gets into the frame buffer of the video card and is displayed on the monitor screen.

Electronic components of the video card

Now, when it became clear how the process of building a three-dimensional image occurs, you can list the technical characteristics of the video card components that determine the speed of the process. The main components of the video card are graphic processor (GPU - Graphics Processing Unit) and video memory.

Graphic processor

One of the main characteristics of this component (as well as the Central PC processor) is the clock frequency. All other things being equal than it is higher, the faster data is processing, and therefore the number of frames per second (FPS - FRAMES PER Second) increases in computer games. The graph of the graphics processor is important, but not the only one that affects its performance parameter is modern models of production NVIDIA and ATI, which have a comparable level of speed, are characterized by different GPU frequencies.

For NVIDIA adapters possessing high performanceThe GPU clock frequencies from 550 MHz to 675 MHz are characterized. The frequency of the graphics processor is less than 500 MHz have "middling" and cheap low-performance cards.
At the same time, the GPU "top" ATI maps have frequencies from 600 to 800 MHz, and even the cheapest video adapter frequency of the graphics processor does not fall below 500 MHz.

However, despite the fact that NVIDIA graphics processors have a lower frequency than the GPU developed by ATI, they provide at least the same level of productivity, and often - and higher. The fact is that no less value than the clock frequency have other GPU characteristics.

1. The number of texture modules (TMU - Texture Mapping Units) - elements of the graphics processor that perform the imposition of textures on triangles. From the number of TMU directly depends the speed of building a three-dimensional scene.
2. The number of render conveyors (ROP - Render Output Pipeline) - blocks performing "service" functions (a couple of examples, pls). In modern ROP graphics processors, as a rule, less than texture modules, and it limits the overall texture rate. For example, the NVIDIA GeForce 8800 GTX video card chip has 32 "textural" and 24 ROP. The processor video card ATI Radeon. HD 3870 Only 16 texture models and 16 ROP.

The performance of textural modules is expressed in such a magnitude as Filreite - the texturing rate measured in Texes in a second. The GeForce 8800 GTX video card has a Filreite in 18.4 billion Tex / s. But the more objective indicator is the Filreite measured in pixels, as it reflects the speed of the ROP. GeForce 8800 GTX This value is 13.8 billion pixes. / S.
3. The number of shader blocks (shader processors), which - as follows from the name - are engaged in processing pixel and vertex shaders. Modern games actively use shaders, so the number of shader blocks is crucial to determine performance.

Not so long ago, graphic processors had separate modules for performing pixel and vertex shaders. The NVIDIA GEFORCE 8000 series and ATI Radeon HD 2000 adapters are first switched to a unified shader architecture. Graphic processors of these cards have blocks capable of processing both pixel and vertex shaders - universal shader processors (streaming processors). This approach allows you to fully use the computing resources of the chip in any ratio of pixel and vertex calculations in the game code. In addition, in modern graphics processors, the shader blocks often operate at a frequency exceeding the GPU clock frequency (for example, the GeForce 8800 GTX, this frequency is 1350 MHz against the "common" 575 MHz).

We draw your attention to the fact that NVIDIA and ATI in different ways consider the number of shader processors in their chips. For example, Radeon HD 3870 has 320 such blocks, and the GeForce 8800 GTX is only 128. In fact, ATI indicates instead of whole shader processors their components. Each shader processor contains five components, so that the total number of shader blocks at the Radeon HD 3870 is only 64, so this video card works more slowly than the GeForce 8800 GTX.

Memory video card

The video memory with respect to the GPU performs the same functions as RAM - in relation to the PC central processor: it stores the entire "building material", which is necessary to create an image - texture, geometric data, shaders programs, etc.

What are the characteristics of video memory affect graphics card performance

1. Volume. Modern games use a huge amount of textures with high resolutionAnd for their placement requires the corresponding volume of video memory. The bulk of today's "top" video adapters and medium price range cards is provided with 512 MB of memory, which cannot be increased subsequently. Cheaper video cards are equipped with a smaller amount of memory, it is not enough for modern games.

In the event of a shortage of memory, the graphics processor is forced to constantly load textures from random access memory PC, the connection with which is carried out much slower, as a result, performance may noticeably decrease. On the other hand, an overly large amount of memory may not give any speed increase, since the additional "place" simply will not be used. Buy a video adapter with 1 GB of memory makes sense only if it belongs to "top" products (ATI Radeon HD 4870 video cards, NVIDIA GeForce 9800, as well as new cards Series GeForce GTX. 200).

2. Frequency. This parameter in modern video cards can vary from 800 to 3200 MHz and depends primarily on the type of memory used by the microcircuit. DDR 2 chips can provide operating frequency within 800 MHz and are used only in the cheapest graphics adapters. The memory GDDR 3 and GDDR 4 increases the frequency range up to 2400 MHz. The newest graphics cards ATI Radeon HD 4870 use GDDR-5 memory with a fantastic frequency - 3200 MHz.

The frequency of memory, as well as the frequency of the graphics processor, has a big impact on the performance of the video card in games, especially when using full-screen smoothing. All other things being equal, the greater the frequency of memory, the higher the speed, because The graphics processor will be less "idle" waiting for data receipt. The memory frequency of 1800 MHz is the lower boundary separating high-performance maps from less quick.

3. The battery of the video memory bus is much more effectively affected by the overall card performance than the memory frequency. It shows how much data can pass the memory for one clock. Accordingly, a two-time increase in the discharge of the memory bus is equivalent to doubling its clock frequency. The bulk of modern video cards have a 256-bit memory bus. Reducing the bit to 128 or, especially, up to 64 bits causes a strong blow to speed. On the other hand, in the most expensive video cards, the tire can be "extended" to 512 bits (until only the newest GeForce GTX 280 boasts, which turns out to be quite by the way, taking into account the power of their graphics processors.

Where to find information on the technical characteristics of the video card

If a graphic map. It has some outstanding parameters (high clock frequency of the processor and memory, its volume), then they are usually indicated directly on the box. But the most complete specifications of the video adapter and GPU, on which they are based, can only be found on the Internet. General information is postponed on corporate graphics processor manufacturers: NVIDIA (www.nvidia.ru) and ATI (www.ati.amd.com/en). Details can be found on unofficial websites dedicated to video cards - www.nvworld.ru and www.radeon.ru. A good help will be the electronic encyclopedia of Wikipedia (www.ru.wikipedia.org). Users buying a card with an overclocking sight can take advantage of the resource www.Orclockers.ru.

Simultaneous use of two video cards

To receive you need to maximum productivity, You can install two video cards in your computer. Manufacturers provided for this appropriate technologies - SLI (Scalable Link Interface, used by NVIDIA cards) and CrossFire (ATI Development). In order to take advantage of them, the motherboard should not only have two PCI-E slots for video cards, but also support one of these technologies. Many "motherboards" on intel chipsets Can use ATI cards in CrossFire mode, but to combine two (or even three!) The video cards of NVIDIA can only boards on the chipsets of the same company. If the motherboard does not have support for these technologies, two video cards will be able to work with it, but only one will be used in games, and the second will only provide an opportunity to display an image to a pair of additional monitors.
Note that the use of two video cards does not lead to doubling performance. The average result that is worth calculating - 50% of the speed of speed. In addition, the entire potential of the tandem will be disclosed only when using a powerful central processor and a high-resolution monitor.

What is shaders

Shaders are firmware present in the game code, with which you can change the process of building a virtual scene, opening the possibilities unattainable when using traditional 3D rendering tools. Modern game graphics without shaders are unthinkable.

The vertex shaders change the geometry of three-dimensional objects, thanks to which it is possible to implement the natural animation of complex models of game characters, physically correct deformation of objects or real waves on water. Pixel shaders are used to change the color of pixels and allow you to create such effects such as realistic circles and ripples on water, complex lighting and relief surfaces. In addition, with the help of pixel shaders, a post-processing of frame is carried out: all sorts of "cinematic" effects of the blurring of moving objects, super-light light, etc.

There are several versions of the shader model (Shader Model). All modern video cards support pixel and vertex shaders version 4.0, ensuring compared to the previous one - the third version of the more realistic effects. Shader Model 4.0 is supported by the DirectX 10 API, which works exclusively in windows Environment Vista. In addition, computer games themselves must be "sharpened" under DirectX 10.

Do you need agp video card with an old system

If the "motherboard" of your PC is equipped with a port of AGP, the capabilities of the upgrade video card are strongly limited. The maximum that can afford the owner of such a system is the Radeon HD 3850 AMD (ATI) series video cards.

According to modern standards, they have a performance below average. In addition, the overwhelming majority of motherboards with AGP interface support is intended for obsolete intel processors Pentium 4 and AMD Athlon XP, so that the overall speed of the system will still be high enough for modern three-dimensional graphics. Just on motherboards for aMD processors Ahtlon 64 with Socket 939 connector It is worth installing new video cards with AGP port. In all other cases, it is better to buy new computer with pCI-E interface, DDR 2 memory (or DDR 3) and modern CPU.

Tags material:graphic Map, Video, Map, Accelerator, Graphics

Task Manager Windows 10. Contains detailed monitoring tools graphic processor (GPU). You can view the use of each application and a system-wide graphics processor, and Microsoft. promises that indicators task Manager Will be more accurate than third-party utilities.

How it works

These features graphic processor were added to update Fall Creators for Windows 10 also known as Windows 10 Version 1709 . If you use Windows 7, 8 or more old version Windows 10, you will not see these tools in your Task Manager.

Windows uses newer functions in Windows Display Driver Model to extract information directly from graphic processor (VIDSCH) and the video memory manager (VIDMM) in the WDDM graphic core, which are responsible for the actual distribution of resources. It shows very accurate data regardless of what api Apps Use to access GPU-Microsoft DirectX, OpenGL, Vulkan, OpenCl, Nvidia Cuda, AMD Mantle or something else.

That is why B. task Manager Only systems with compatible WDDM 2.0 are displayed. graphics processors . If you do not see this, your graphics processor of your system probably uses an older type of driver.

You can check which version of WDDM is used by your driver. GPUby clicking the Windows + R button by typing in the "dxdiag" field, and then press "ENTER" to open the tool " DirectX diagnostic tool" Click the "Screen" tab and look at the "Model" to the right in the "Drivers" section. If you see the WDDM 2.x driver here, your system is compatible. If you see the WDDM 1.x driver here, your GPU not compatible.

How to view graphics processor performance

This information is available in task Manager Although by default it is hidden. To open it, open Task ManagerBy clicking right-click mice on any empty place on the taskbar and choosing " Task Manager"Or pressing Ctrl + Shift + Esc on the keyboard.

Press the "More details" button at the bottom of the window " Task Manager"If you see a standard simple look.

If a GPU is not displayed in Task Manager , at full screen mode on the tab " Processes"Right-click any column header, and then turn on the option" Graphic processor " This will add column graphic processor which allows you to see the percentage of resources graphic processor used by each application.

You can also enable the option " Core graphics processor"To see which graphics processor uses the application.

General use GPU All applications in your system are displayed at the top of the column. Graphic processor. Click Column GPUto sort the list and see which applications use your GPU Most of all this moment.

Number in column Graphic processor - This is the highest use that the application uses for all engines. For example, if an application uses 50% 3D GPU engine and 2% video decoding GPU engine, you will simply see that the number 50% is displayed in the GPU column.

In the column " Core graphics processor»Each application is displayed. It shows you what physical GPU. And what engine uses the application, for example, whether it uses a 3D engine or video decoding mechanism. You can determine which graphics processor corresponds to a specific indicator by checking the " Performance", We will talk about in the next section.

How to view using video memory applications

If you are interested in how much video memory is used by the application, you need to go to the "Details" tab in the Task Manager. On the "Details" tab, right-click any column header and select "Select columns". Scroll down and turn on the columns " Graphic processor », « Core graphics processor », « "And" " The first two are also available on the Processes tab, but the last two memory options are available only in the "Details" panel.

Column " Selected graphics processor memory "Shows how much memory app uses on your GPU. If your PC is installed discrete video Card NVIDIA or AMD, then this is part of his VRAM, that is, how much physical memory on your video card uses the application. If you have built-in graphic processor Part of your usual system memory is reserved exclusively for your graphic equipment. This shows which part of the reserved memory is used by the application.

Windows Also allows applications to store some data in the conventional DRAM system memory. Column " Shared graphics processor memory »Shows how much memory the application uses currently for video devices from the usual computer system RAM.

You can click any of the columns to sort over them and see which application uses the most resources. For example, to view applications that use the largest video memory on your graphics processor, click the Column " Selected graphics processor memory ».

How to track the use of a general GPU resource

To track overall resource use statistics GPU, go to the " Performance"And look at" Graphic processor"Down on the sidebar. If you have several graphic processors on your computer, here you will see several options. GPU.

If you have several related graphics processors - using such a feature as NVIDIA SLI or AMD Crossfire, you will see them identified "#" in their name.

Windows Displays use GPU in real time. Default task Manager Attempts to display the most interesting four engines in accordance with what happens in your system. For example, you will see different graphs depending on whether you play 3D games or encode video. However, you can click any of the names over the charts and select any of the other available engines.

The name is yours GPU Also displayed on the sidebar and at the top of this window, which makes it easy to check which graphic equipment is installed on your PC.

You will also see graphs of using dedicated and shared memory. GPU. Using shared memory GPU refers to which part of the general memory of the system is used for tasks GPU. This memory can be used both for conventional system tasks and video recordings.

At the bottom of the window, you will see information, such as the version number of the installed video driver, the development date and physical location GPU on your system.

If you want to view this information in a smaller window, which is easier to leave on the screen, double-click somewhere inside the graphics processor screen or right-click anywhere inside it and select the " Graphic summary" You can deploy the window by double-clicking on the panel or right-click on it and removing the checkbox. Graphic summary».

You can also right-click on the schedule and select "Change Chart"\u003e "One kernel" to view only one schedule engine GPU.

To make this window constantly displayed on your screen, click "Parameters"\u003e On top of the other windows».

Double-click inside the panel GPU Once again, and you will have a minimal window that you can locate anywhere on the screen.

Not many users know that video cards can perform much more than just display a picture on the monitor. Using CUDA, Stream and other similar technologies, you can significantly raise the performance of the computer, having lost your own calculations. Below will be described the principle of operation.

To display continuous frames in some modern game, the computer requires good performance. It should be assumed that modern performance video cards correspond to fresh versions processors.

It is worth noting that when the video adapter is idle and does not perform image processing, its capabilities remain unclaimed. So that there is no such downtime and one could take some responsibilities for it, which will reduce the load on the processor, it is necessary to apply special options for accelerating the computer. Below will be detailed instructions On the principles of this technology, which can increase PC performance.

How does the video payment increase the speed of the computer?

Use the capabilities of video cards can only special applications. These programs can be combined with video card and use one of the 4 physical acceleration technologies.

CUDA. This development was created by NVIDIA. This technology can be used to carry out complex computing manipulations and editing video and pictures.

Stream. This mechanical acceleration technology is similar to the first, but developed by the manufacturer of AMD video adapters.
Both of these technologies are supported by all Operations, except Mac OS, and are used only with video cards of a suitable manufacturer. The creators are forced to carry out additional work so that the video cards of both developers can increase the speed of their applications. Below are the technologies that are capable of working with boards of both manufacturers.

OpenCl. This technology has been released Apple Corporation in 2008 and is supported by all Operations and any software. However, today there are no applications to speed up a computer using this technology. In addition, to increase OpenCL productivity significantly behind the first two technologies.

DirectCompute. This technology has been built into Microsoft in DirectX 11. But it is capable of working only on the Windows 7 and Vista Operations, and then with a small application package.

What increase in performance provides a video card?

The increase directly depends on the graphics adapter and the performance of the other components of the computer. An increase in productivity is set by utilities and operations performed. On the modern average PC, an increase in the speed of transformation of high-quality video can reach up to 20 times. But the editing by filters and special effects photograph can accelerate three hundred times.

What affects the high productivity of CUDA and similar technologies?

CPU on the motherboard When performing complex tasks, it initially separates the process to a slightly smaller, and after it performs their sequential processing. The resulting intermediate result is placed in a small, but fast processor memory. When the memory departments are overwhelmed, the files are moved to the cache memory, which is also located in the processor. But the exchange of information between the processor and the RAM requires quite a long time, so the speed is not entirely high.

Video cards can sometimes carry out such manipulations much faster. This may affect several circumstances. One of them are parallel computing. If necessary, conduct several similar manipulations, some of them can be carried out by a graphic module in conjunction with the processor.

For example, when working with video or pictures, the utility must be changed a huge number of pixels, and at the same time using repeating methods. Especially for this, the graphic adapter has hundreds of small processors that wear streaming names.

In addition, it is necessary fast access to memory. By analogy with central processes, graphic adapters also have their intermediate memory and RAM. But in this case they have many registers high-speed memorywhich significantly increases the speed of calculations.

What number of streaming CPU have video cards?

This affects the processor model. For example, GeForse GTX 590 has two FERMI modules, each of which has 512 streaming CPUs. One of the most powerful video payments from AMD - Radeon HD 6990 is also equipped with a pair of modules, each of which is 1536 processors. But with all this, HD 6990 significantly loses GTX 590 speed.

How to run CUDA or STREAM?

Do not run anything, since technology is an element of the hardware of video cards. After the graphics adapter driver set the application that supports some kind of technology, then an increase in the speed of the computer will automatically occur. To get full performance, you need to install the latest version of the driver.
It is worth noting that the AMD video card users need to download and install the AMD Media Codec Package set.

Why not all utilities work with these technologies?

Until the OpenCL is widespread, creators software It is necessary to adjust each application to work with NVIDIA and AMD video plates. But at the same time, not every manufacturer will go to additional costs.

In addition, not all applications have the ability to provide a constant stream of simple computing operations that can occur in parallel. It can work perfectly with video editing programs and graphics. For posts or text editors These technologies will not help much.

Super PC

For example, the Chinese TIANHE-1A PC has 7168 NVIDIA graphic modules that support excellent performance. It takes 2.5 trillion calculations per second. This computer consumes 4 megawatta energy. So much electricity spends the town with five thousand people of the population.

Is the graphic adapter be able to replace the central?

This replacement is impossible. The device of these processors is completely different. The CPU is a universal computing unit that has the ability to process and send information to other PC elements. In turn, video cards are narrowly controlled devices, despite the fact that they perform a small number of operations, but at the same time at high speed.

What will happen in the future: universal chips

To increase CPU performance, Intel and AMD corporations constantly add kernels to their processors. In addition, they constantly add new technologies that are able to increase the efficiency of computing operations and the possibility of parallel information processing.

Compared to the central processors, the video cards already have a large number of simple cores that are capable of performing complex calculations very quickly.

But it turns out that the initial differences in the principles of the video card and the CPU are gradually erased. Therefore, the development of a universal chip is very logical. Today, computer users can use the full potential of the video card without expensive graphic chips.

Modern processors from leading developers, at the moment can demonstrate the ability to connect the graphics adapter and CPU and work as one universal computing unit.

In any of the chips of the CPU kernel and the video card are placed on a single crystal. This provides an opportunity to quickly divide the computational manipulations between the cores. These technologies used are the name of Intel Quick Sync and AMD Arr. At this time, there are already separate applications that use such technology.

In general, it's all that you need to know about the differences in the central processor and the video card. As can be seen from the written, the graphics processor is capable of performing some central operations, especially modern computers with powerful video cards.

We all know that the video card and processor have several different tasks, but do you know what they differ from each other in the internal structure? Like CPU (eng. - central Processing Unit.) and gpu (eng. - graphics Processing Unit.) are processors, and there are many common between them, but they were constructed to perform various tasks. You will learn more about this from this article.

CPU.

The main task of the CPU, if we talk simple wordsThis is the execution of the instruction chain for the shortest possible time. The CPU is designed in such a way as to perform several such chains simultaneously or split one stream of instructions into several and, after executing them separately, merge them again into one, in the correct order. Each instruction in the stream depends on the following after it, and that is why there are so few executive blocks in CPU, and the entire emphasis is on the speed of execution and reduce downtime, which is achieved using cache memory and conveyor.

GPU

The main function of GPU - rendering 3D graphics and visual effects, therefore, everything is a bit simpler: it needs to be obtained at the entrance of the polygon, and after the necessary mathematical and logical operations over them are necessary, to output the coordinates of pixels. In fact, the GPU work comes down to operating over a huge number of independent tasks, therefore, it contains a large amount of memory, but not as fast as in the CPU, and a huge number of executive blocks: in modern GPUs are 2048 and more, at that time As a CPU, their number can reach 48, but most often their number lies in the range of 2-8.

Main differences

CPU differs from GPU primarily ways to access memory. In the GPU it is connected and easily predictable - if the texture texture is read from the memory, then after a while there will be a queue and neighboring texels. With the recording of a similar situation, the pixel is written to the framebufer, and after several clocks will be recorded located next to it. Also a graphical processor, in contrast to universal processors, simply does not need large-size cache memory, and only 128-256 kilobytes are required for textures. In addition, the video cards apply more fast memoryand as a result of GPUs are available at times big bandwidthThat is also very important for parallel calculations that operate with huge data streams.

There are many differences and in support of multithreading: CPU executes 1 – 2 calculation flows per processor core, and the GPU can support several thousand streams on each multiprocessor, which in the chip several pieces! And if the switching from one stream to another for the CPU is worth hundreds of clocks, then the GPU switches several streams per tact.

In CPU, most of the chip area is occupied by the command buffers, the hardware prediction of branching and huge amounts of cache, and in the GPU most of the area are occupied by executive blocks. The above-described device is schematically shown below:

Difference in the speed of calculations

If the CPU is a kind of "boss", which makes decisions in accordance with the instructions of the program, then the GPU is a "working" that produces a huge amount of single-type computing. It turns out that if you serve on the GPU independent simplest mathematical tasks, it will cope much faster than the central processor. This difference successfully enjoy miners bitcoins.

Mining Bitcoin.

The essence of mining is that computers located at different points of the Earth solve mathematical tasks resulting from bitcoins. All Bitcoin Translations on a chain are transferred to Majneram, whose work is to choose from millions of combinations a single hash, suitable for all new transactions and a secret key, which will provide Mainer to obtain a reward in 25 bitcoins at a time. Since the calculation rate directly depends on the number of executive blocks, it turns out that the GPU is significantly better suited for execution this type Tasks rather than CPU. The greater the number of calculations produced, the higher the chance to get bitcoins. The point even reached the construction of whole farms from the video cards.