How do 3D graphics cards work

Graphics cards

The graphics card calculates visual data and outputs it via an interface to which a screen is connected. The user interface of the operating system and application programs is then displayed on the screen.
Graphics cards and their functions largely determine the system performance and properties of a computer. The performance of the graphics card is mainly relevant for computer games and extensive graphic displays. In complex game scenes, the graphics card calculates detailed objects and the right lighting. While the main processor does general calculations and, for example, calculates the behavior of game characters, the graphics card takes over the calculation of physical phenomena, explosions and collapsing buildings, for example. If the screen display jerks with these movements or shifts, then the graphics card, especially the graphics processor (GPU), is not powerful enough with the currently selected resolution and level of detail.

Graphics processing is increasingly being relocated from the graphics card to the main processor. However, that doesn't mean that the days of graphics cards are over. It is more the case that the graphics functions, which are already integrated "onboard" in the chipset, have to be integrated into the processor right away. Consequently, the graphics must also be output from the processor.

Contrary to what you might think, image editing programs benefit above all from a fast CPU and a lot of RAM. The graphics card or graphics processor mainly improves the speed of the screen display. For example, when moving windows and picture elements, continuously zooming and rotating the drawing area. The graphics card only supports the computer system directly in processing image and video data in certain cases. For example when playing and transcoding videos.

Graphics card with TV and radio tuner Graphics card with dual head DVIStandard graphics card

Structure of a graphics card

A graphics card consists of the following mechanical and electronic parts:

  • cover
  • Fan
  • Heat sink
  • Graphics processor
  • Storage
  • circuit board
  • Slot bracket
  • further coverage

The most important task of a graphics card is to output an image signal for screens. To do this, the main processor transfers the data to the graphics card.
Due to the high volume of data between the processor and graphics card, graphics cards are equipped with their own processor. It is supposed to relieve the main processor with arithmetic operations running in parallel. The heart of the graphics card is the graphics processor (GPU), which determines the essential performance characteristics of a graphics card. There are also graphics cards that have two GPUs and that have to be cooled with a heat sink or even a fan due to the high power dissipation.
In order to be able to process the enormous amounts of data, graphics cards have their own main memory and a special memory connection with high clock rates. The graphics memory can be as large as normal RAM. A memory controller controls access to the graphics memory.
The RAM-DAC is the output stage that converts the digital information from the image memory into an analog signal for the VGA connection. The performance of the RAM-DAC is indicated by its speed in MHz. The higher this value, the higher resolutions and color depths (number of colors) are possible. Since digital signals are usually sent to the displays today, the RAM-DAC has lost its importance. If the graphics card has a VGA connection or outputs VGA signals via DVI, the RAM-DAC still exists.
A bus interface ensures that the graphics card is connected to the computer system. The bus interface uses PEG (PCIe) to transfer the data with the chipset or the processor. Older graphics cards use AGP or PCI.

GPU - Graphic Processing Unit

The GPU is nothing more than a co-processor designed for 3D calculations. The realistic and spatial representation of objects with shadows and moving scenes is the main work of a graphics card. Special physics engines simulate physical laws from mechanics in 3D games or applications. For example, it is possible for cars to suffer realistic damage after a collision. A normal CPU is far too slow for such calculations.
The 3D performance of a GPU is mainly achieved through parallelization and the fact that typical graphics processing functions are implemented in hardware. For special tasks, GPUs are much faster than the most expensive processors. This includes, for example, HD video transcoding, which can make a normal processor sweat quite a bit.
The GPU sits either in the chipset, in the processor or on a graphics card.


A shader is a small processor inside the GPU. A GPU has several hundred shaders. The shaders are optimized for different functions or for specific calculations. In older GPUs it was common to differentiate between special processing units for vertex and pixel calculations. However, the units were then used rather unevenly. Since DirectX-10 there have only been unified shaders that act as vertex or pixel shaders as required.
The vertex shader calculates the position of an object and also takes care of the correct lighting. Once the vertex shaders have done their calculations, the object is covered with color or structures. The pixel shaders are responsible for this.
The structures are also referred to as textures. A texture is a digital tile that is glued to an object. It is used several times in an image or in a sequence of images. These textures must be loaded into the graphics memory before the calculation. Most of the graphics memory is occupied by textures. If there is not enough memory, the graphics card has to reload textures from the hard drive over and over again. This can cause image jerks. The larger the memory, the more textures can be preloaded.

Image enhancement

Graphics processors have additional functions to make 3D representations look more realistic. Anisotropic filtering sharpens the textures of surfaces. They appear clearer and bumps seem to stand out.
Image content is always calculated by the GPU with a higher resolution and then downscaled to the resolution of the monitor. This creates ugly steps on the edges. With antialiasing (edge ​​smoothing) the image appears softer and the stairs have almost disappeared.
Tessellation increases the level of detail in images by the graphics card adding additional points in the image and modeling even finer structures.

HDR - High Dynamic Range

High Dynamic Range, or HDR for short, describes images with a high contrast range. In concrete terms, this means that the images have much more contrast and stronger colors. For example, a very dark black and a very light white.
This real HDR should not be confused with the 3D effect of the same name. It just simulates this impression.

3D acceleration

So that every game manufacturer does not have to reprogram the 3D functions of the individual graphics card manufacturers, graphic interfaces, so-called application program interfaces (APIs), have been developed. These APIs make the software largely independent of the graphics card. They provide functions for calculating and building up the image. There are two of these interfaces. One is DirectX from Microsoft and the other is OpenGL. OpenGL is an open standard that is included in almost all operating systems.


3D interface with a low-level API that developers can use to program close to the hardware in order to better utilize modern processors and graphics chips. Better than DirectX or OpenGL.
In addition, Vulkan runs on both desktop and mobile hardware and any operating system.

VR - Virtual Reality

Virtual reality in itself is nothing new. In the past, this buzzword was buzzing around the country from time to time. Virtual reality was mostly a gimmick. The computing power was not big enough for real VR and so the applications also fell short of expectations. In addition, the presentation on a conventional monitor was anything but virtual reality.
That has changed with VR glasses. However, here, too, it is not certain whether VR will find its way to a wide audience in this form. Because the hurdles for VR are high. The glasses are expensive, require a fast computer, compelling applications are in short supply, and real-life applicability is limited. In principle, the application is reduced to gaming and perhaps training on virtual objects. The whole VR industry is desperately waiting for the killer app to justify the investment and the high price. Virtual Reality is now very impressive, but also more entertainment and gaming than practical benefits.

How does a graphics card produce an image?

The display or calculation of an image begins in the software. For example in a game or image editing software. Commands for displaying image information are also processed within the program code. These commands are passed to the operating system. For example, DirectX or OpenGL are available for further processing. A graphics card driver then translates the DirectX commands, which are then executed by the graphics card. Each graphics card needs its own driver for each operating system in order to work optimally. Operating systems also provide standard drivers. But they only support rudimentary functions. A driver serves as an interface between hardware and software.
The pixels are processed one after the other in pipelines. In the best case scenario, several pipelines are available. Then several processing steps can run in parallel and the image structure can be calculated more quickly. Within the pipeline there are shaders that can give the pixel areas colors or geometric shapes. Shaders come into action primarily when graphic effects are required.
When a picture is complete, the graphics card forgets all calculations and starts the calculations all over again with the next picture. Only the loaded vertex buffers, textures and shader programs are retained when the GPU needs them again in the next image. Only new data is then added. For example, when new objects or details appear in the picture.

Choosing a graphics card

A distinction is made between graphics cards with regard to their 3D or gaming suitability, their performance and variety of functions. The graphics card is to be selected according to the requirements of the application. A distinction is made between graphics cards according to their price in entry-level, mid-range, performance and high-end models. Depending on the price, graphics cards differ in their 3D performance and features. Typical performance characteristics are clock frequency, number of shaders, memory clock and memory bandwidth.

With an entry-level graphics card, gaming suitability is limited, performance is poor and the range of functions is limited. Simple office and internet applications get by with an inexpensive graphics card or onboard solution. With a high-end graphics card, the suitability for games is unbeatable, the performance is enormous and the variety of functions leaves nothing to be desired. (Action) games in particular inevitably require a powerful graphics card.

Until a few years ago you always had to install the fastest and therefore most expensive graphics card in your PC if you wanted to play the latest computer games. But since many computer games have also been developed for the XBox by Microsoft and Playstation by Sony, the requirements have been decreasing. This is because the game consoles remain in operation for longer and cannot be upgraded afterwards. Therefore, the developers orient themselves to the hardware specifications of the game consoles and adapt these games as multi-platform games to the PC. As a result, even entry-level graphics cards for the PC are many times more powerful than the graphics function of a game console.
For 120 to 150 euros you get a graphics card that works very quietly under Windows and whose performance is sufficient to play most multi-platform games smoothly in full HD resolution and to play them without any problems. In addition, these graphics cards are also suitable for playing back films from Blu-ray Disc.

 BeginnersMiddle classperformanceHigh end
priceup to 75 EURup to 150 EURup to 300 EURup to 1000 EUR
Number of shadersup to 160up to 800until 2000up to 4000
Memory size512 Mbytes1 GB2 GB4 GB
typical 3D
Power consumption
up to 50 wattsup to 90 wattsup to 170 wattsup to 300 watts
or more
power adaptermin. 300 wattsmin. 350 wattsmin. 450 wattsmin. 500 watts
or more

Processor graphics / chipset graphics / onboard graphics

In addition to the graphics card for installation in a computer, there are also onboard versions. Either as an independent chip on the motherboard or integrated into the chipset or processor. In order to keep the power consumption, complexity and costs of the chipset low, the performance of the graphics chip in particular is severely restricted. In addition, part of the main memory is used by the onboard graphics (shared memory). So the full working memory is no longer available. As a rule, the onboard graphics chips lag behind the graphics cards in terms of performance.
Motherboards with integrated graphics processors are characterized by the very poor quality of the signals at the VGA interface. It is generally advisable to choose the digital DVI interface for onboard graphics.
An integrated graphics unit is completely sufficient for working with Office, surfing the Internet and reading e-mail. Graphics performance does not play a role in image and video editing either. A fast main processor and a lot of RAM are required here. Only a few image processing programs use the GPU for their calculations.
The internal graphics unit is suitable for accelerating the image build-up. However, this does not require a particularly fast graphics unit. All DirectX-compatible onboard GPUs can do this. HD video decoding is also part of the standard tasks of an onboard graphics unit. There are differences in quality and efficiency here. These depend on the settings of the player software and the graphics card driver.

Entry-level graphics cards (low-end / low-cost)

The entry-level graphics cards often include formerly modern graphics cards that have now been replaced by newer models. These are graphics cards that run absolutely flawlessly. The price is less than 50 euros. You are at the lower end of the price and performance segment. They impress less with a good performance, but more with the added value of driver and software equipment compared to the integrated chipset graphics. Compared to mid-range or high-end graphics cards, they have fewer processing units, lower chip and memory clock rates and a slower memory connection. They are too slow for many 3D games. Some games can be played with reduced resolution and lower details. However, the fun of the game suffers from this.
Due to the low power consumption, they get by with passive cooling (heat sink).

Mid-range graphics cards

Mid-range graphics cards can show most computer games smoothly. In the case of very complex games, however, the maximum depth of detail should not be adjustable. This means that you have to accept certain compromises in the representation of light and shadow. But games only become interesting with mid-range graphics cards. Even demanding games can be persuaded to run smoothly by tweaking the graphics options.

Performance graphics cards

Performance graphics cards leave nothing to be desired. But they are only worthwhile for computer gamers who can also accept higher power consumption and thus higher power consumption.

High-end graphics cards

High-end graphics cards are only worthwhile for demanding gamers. These are graphics cards that have functions that are hardly supported by any current computer game. Most of the time, the games have to be programmed for the special functions.
With the enormous 3D performance, the power consumption also increases. Because a PCIe slot does not deliver more than 75 watts, high-end graphics cards must be supplied with power via one or two additional power connections (6-pin, 75 watts each). A 450 watt power supply unit is required for each external power connection. You have to take into account that the entire system also consumes electricity. In order to dissipate the waste heat from these graphics cards, cooling structures are attached to the graphics cards that occupy two housing slots.
Models with two processors (GPUs) do not double the performance. At best, an increase of 50 to 70 percent can be achieved. This also applies to the AMD Crossfire or Nvidia SLI operation of two graphics cards. Also, not all games benefit from two or more GPUs. Their use is therefore only useful to a limited extent.

Multi-GPU architecture

Multi-GPU architectures are used to increase the graphics performance of a PC. It is popular to install two or more graphics cards in a PC. A special chipset and one or more power supplies with a total of 800 to 1000 watts are required for this.The SLI or Crossfire interface is used to connect two graphics cards.

Compare the performance of graphics cards

Graphics cards are often compared on the basis of their equipment and performance characteristics. The clock frequencies, number of shaders, memory clock, memory bandwidth and graphics memory are compared. Unfortunately, such a comparison says little about the performance of a graphics card. It would be nice if these performance features were coordinated in such a way that the best possible performance comes out of them. But you don't know that exactly in advance.
In connection with graphics cards one often finds the indication FPS, which means frames per second. FPS indicates how many screen displays a graphics card can produce per second. One also speaks of the frame rate. That means how often the image is recalculated and built up. FPS values ​​that are too low result in a jerky picture. In general, you can say that a smooth picture is displayed from a frame rate of 30 fps. For some games it should be a little more.

In principle, the frame rate says what a graphics card is capable of. In order for the frame rate to be meaningful, you usually need to specify which resolution a certain frame rate applies to. Because the greater the resolution, the more memory and computing power a screen display consumes. Only the combination of resolution and frame rate allows comparisons between graphics cards.
The frame rate is usually only of interest to computer gamers. For them, for example, it is important that the realistic representation in first-person shooters runs smoothly.

Adaptive-Sync / G-Sync

Adaptive-Sync is a standard of the VESA (Video Electronics Standards Association) and part of DisplayPort from version 1.2a, with which displays adapt their refresh rate dynamically to the image output of the graphics card. The idea behind this is to only change the image display when the graphics card has delivered a new image. It doesn't matter how fast the frame rate is in frames per second (frames per second, fps).

Adaptive-Sync is necessary because the frame rate collapses with increasing speed and image details, which causes the image to jerk on the display. Gamers in particular know the problem. Adaptive-Sync is based on AMD's FreeSync. For this, both the graphics card and the display must be compatible with each other.
In addition to Adaptive-Sync, there is also a solution from nVidia with the proprietary G-Sync, which does the same in principle.

Overclocking a graphics card

Overclocking graphics cards is no problem with a few clicks with the help of the driver and appropriate tools from the manufacturer. Nevertheless, extreme caution is required when overclocking. The load on the voltage converters and the temperature of the graphics chips increase at higher clock frequencies. In the worst case, the graphics card breaks.
Overclocking graphics cards brings a performance gain at most a few percent, but also many disadvantages. An overclocked PC can quickly become unstable. Image errors, calculation errors and system crashes occur.
An alternative are graphics cards that the manufacturer has overclocked by default. They run stably and are also fast.

Interfaces for graphics cards

In the beginning graphics cards were operated via the internal system bus. With increasing resolution, the demand for multimedia and 3D display, the amount of data that had to be transported to the graphics card increased. In order to be able to cope with the increasing amounts of data between the processor, main memory and graphics card, Intel introduced the AGP slot for graphics cards. This slot was only intended for graphics cards. A few years later, the AGP was replaced along with the PCI by the PCI Express (PCIe). Today's graphics cards are PEG graphics cards (PCIe).

Connections for screen / monitor

If the display is poor, then the screen is almost always responsible. If the screen is connected to the graphics card analog, i.e. via the VGA, then the signal quality of the VGA can be responsible for the poor picture. Although one might assume that this shouldn't be a problem today, some graphics cards have poor signal quality. The VGA connections in particular have these problems. You should preferably use the DVI or DisplayPort connection.


Although DVI is a digital interface for high-quality image and video signals, the introduction of new interfaces makes perfect sense. Because DVI is not suitable for high resolutions. The DisplayPort is intended to replace DVI (Digital Visual Interface) in PCs and HDMI (High Definition Multimedia Interface) in entertainment electronics.
The DisplayPort is electrically compatible with the existing digital inputs DVI and HDMI. It also supports the HDCP copy protection mechanism and also has its own protocol called DPCP (Display Port Content Protection).

Display devices

Other related topics:

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage devices, drives and important hardware components.

I want that!

Everything you need to know about computer technology.

Computer technology primer

The computer technology primer is a book about the basics of computer technology, processor technology, semiconductor memory, interfaces, data storage devices, drives and important hardware components.

I want that!