What is a video card?

Video Card


Video card (also known as a graphics card, graphics card, graphics card, graphics card) - a device that converts a graphic image that is stored as the contents of your computer's memory or the adapter itself, in some form, intended for further output on the screen. Currently this feature is essential and lost in the first place under the graphics card processor sgraficheskim understand device - a graphics accelerator, which is developing a very graphic image.
Usually the video card is the expansion card and inserted into an expansion slot, a universal (PCI-Express, PCI, ISA, VLB, EISA, MCA) or specialized (AGP), but it happens and built-in (integrated) into the motherboard (as in a separate chip and as a part of the north bridge chipset or CPU). In this case the device is, strictly speaking, can not be called a video card.
Modern graphics cards are not limited to a simple conclusion of the image, they have a built-in graphics processor that can perform additional processing, removing this task from the central protsessorakompyutera. For example, all modern video cards Nvidia and AMD (ATi) shall render the graphics pipeline in OpenGL and DirectX hardware. Recently, also has a tendency to use the computing power of GPUs to solve non-graphics tasks (see OpenCL).

History of Video Cards
One of the first graphic cards for IBM PC became the MDA (Monochrome Display Adapter) in 1981. He worked only in text mode with a resolution of 80 × 25 characters (physically 720 × 350 pixels) and supports the five attributes of the text: normal, bright, inverse, underlined and blinking. No color or graphics, he could not pass, and then, what color are the letters defined model of the monitor. They were usually black and white, amber or emerald. The company Hercules in 1982 released a further development of the adapter MDA, video card HGC (Hercules Graphics Controller - Hercules graphics adapter), which had a graphic resolution of 720 × 348 pixels and supports two graphics pages. But it is still not allowed to work with color.
First color video was CGA (Color Graphics Adapter), published by IBM and became the basis for future standards of graphics cards. It could work either in text mode with resolutions of 40 × 25 and 80 × 25 (character matrix - 8 × 8), or graphically with a resolution of 320 × 200 or 640 × 200. In text modes available symbol attributes 256 - 16 characters and 16 colors of background colors (or 8 background colors and flashing attribute), in graphical mode: 320 × 200 was available four palettes of four colors each, the high resolution mode of 640 × 200 was monochrome. In the development of this map appeared EGA (Enhanced Graphics Adapter) - improved graphics, enhanced with up to 64 color palette, and an intermediate buffer.Was improved up to 640 × 350, resulting in added text mode 80 × 43 matrix with 8 × 8 character. For mode 80 × 25 matrix was used most - 8 × 14, at the same time be able to use 16 colors, the color palette has been expanded to 64 colors. Graphics mode is also allowed to use at a resolution of 640 × 350 16 colors from a palette of 64 colors. Compatible with the CGA and MDA.
It is worth noting that interfaces with the monitor all these types of digital video were Cartan, MDA and HGC transmit only light or no light point and an additional signal intensity for an attribute of the text "bright", the same CGA three channels (red, green, blue) passed basic video, and can additionally transmit a signal intensity (A total of 16 colors), EGA has two transmission lines for each primary color, that is, each primary color can be displayed with full brightness, 2 / 3 or 1 / 3 of full brightness, which and gave a total maximum of 64 colors.
Early computer models of the IBM PS / 2, a new graphics card MCGA (Multicolor Graphics Adapter - color graphics adapter).Text resolution was increased to 640x400, which made it possible to use mode with 80x50 matrix, 8x8, and 80x25 mode for the matrix of 8x16. The number of flowers increased up to 262,144 (64 levels of brightness for each color) for compatibility with the EGA in text mode was introduced by the color table, which converts a 64-color EGA space to the color space MCGA.Mode appeared 320x200x256, where each pixel on the screen corresponding byte in the encoded video, no bit planes were not, according to EGA compatibility remained only on text mode, compatibility with the CGA was complete. Because of the enormous amount of brightness primary color used is a need analog color signal, the frequency of horizontal scanning was already 31.5 kHz.
Then IBM went further and made a VGA (Video Graphics Array - an array of graphic video), this extension MCGA, EGA, and is compatible with medium-sized models introduced in the PS / 2.This is a de facto standard for video Cartan since the late 80's.Added: 720x400 resolution text to emulate MDA and graphics mode 640x480 with access via the bit-plane. 640x480 mode is remarkable in that it uses square pixels, ie the ratio of the number of pixels horizontally and vertically, is the standard aspect ratio - 4:3. Then came IBM 8514 / a Permission 640x480x256 and 1024x768x256, and IBM XGA with text mode 132x25 (1056x400) and increased color depth (640x480x65K).
In 1991 came the notion of SVGA (Super VGA - «over» VGA) - VGA expansion with the addition of higher modes and additional services, such as the possibility to put an arbitrary frame rate.The number of simultaneously displayed colors is increased to 65 536 (High Color, 16 bit) and 16,777,216 (True Color, 24-bit), additional text modes. Of support service functions appears VBE (VESA BIOS Extention - BIOS extension standard VESA).SVGA is perceived as a de facto standard for video Cartan somewhere in the middle of 1992, after the adoption of standard assotsiatsieyVESA VBE 1.0. Up to that point almost all video Cartan SVGA been inconsistent.
The graphical user interface that appears in many operating systems, stimulated a new stage in the development of video Cartan. The notion of a "graphics accelerator» (graphics accelerator). This video Kartal, which make the execution of certain graphics functions in hardware. These functions include: moving large blocks of images from one screen to another area (for example, if you move the window), fill image areas, drawing lines, arcs, fonts, hardware cursor, etc. The direct impetus for the development of such specialized devices was that the graphical user interface, of course, easy to use, but its use requires considerable CPU processing resources, and a modern graphics accelerator is just intended to remove him from the lion's share calculations for the final display on the screen.
An example of a home computer not-IBM - ZX Spectrum, has a history of video modes.

Device video card
The modern video card consists of the following:
GPU (Graphics processing unit - a graphical processing unit) - deals with the calculations of the displayed image, freed from this obligation CPU performs calculations for three-dimensional graphics processing commands. Is the basis for a graphics card, that depend on it and the performance capabilities of the device. Modern graphics processors are not much inferior to the complexity of the central processing computer, and often surpass it as the number of transistors, and in computing power thanks to a large number of universal computing units. However, the previous generation GPU architecture usually implies the existence of several blocks of information processing, namely, the processing unit 2D-graphics processing unit 3D-graphics, in turn, are usually divided into a geometric kernel (plus the cache nodes) and ROP units (plus texture cache ), etc.
video controller - responsible for the formation of images in video memory, gives commands to the RAMDAC on the generation of signals for your monitor and can handle requests of the CPU. In addition, there is usually an external bus controller data (eg, PCI or AGP), the internal data bus controller and the controller memory. The width of the internal bus and memory bus are usually larger than the outer (64, 128 or 256 bits to 16 or 32), many video controllers built yet, and RAMDAC. Modern graphics cards (ATI, nVidia) usually have at least two video controllers, operating independently and simultaneously control a single or multiple displays each.
Video memory - serves as a frame buffer that stores the image generated and constantly changing the graphics processor and displayed on the screen (or multiple monitors). In the video are stored as intermediate elements visible on the screen images and other data. Video memory is of several types that differ in access speed and operating frequency. Modern graphics cards are equipped with memory, such as DDR, DDR2, GDDR3, GDDR4 and GDDR5. It should also be borne in mind that in addition to memory, located on the graphics card, modern GPUs are generally used in the work of the total system memory, direct access to the video driver is arranged through the Cartan or AGP bus PCIE. In the case of UMA architecture as part of the system memory used by your computer's memory.
digital-analog converter (DAC, RAMDAC - Random Access Memory Digital-to-Analog Converter) - used to transform the image formed by the video controller in the levels of color intensity, served on an analog monitor. The possible range of color images is determined only by the parameters RAMDAC.Most often, RAMDAC has four main components: the three-analog converter, one for each color channel (red, green, blue - RGB), and SRAM for data storage on gamma correction. Most have a bit DAC 8-bits per channel - is obtained by the 256 levels of brightness for each primary color, for a total of 16.7 million colors (and by the gamma correction is the ability to display 16.7 million colors starting at a much larger color space) . Some have a bit RAMDAC for each channel 10 bit (1024 levels of brightness), which allows you to immediately display more than 1 billion colors, but this possibility is almost never used. To support a second monitor often sets the second DAC. It is worth noting that monitors and projectors that connect to a digital DVI output of video to convert the digital data stream using its own digital to analog converters and the characteristics of the video DAC are independent.
video-ROM (Video ROM) - read-only memory, which is recorded in video-BIOS, screen fonts, office tables, etc. ROM is not used video controller directly - it refers only to the central processor.Stored in the ROM-BIOS video provides initialization and operation of the video card to boot the main operating system and contains system data that can be read and interpreted by the video driver in the process (depending on the method used, the division of responsibility between the driver and the BIOS).Many modern cards are installed electrically alterable ROM (EEPROM, Flash ROM), allowing recycling of video-BIOS by the user through a special program.
Cooling System - is designed to maintain temperature GPU and memory within an acceptable range.
The correct and fully functional work of contemporary graphic adapter is provided with the video driver - a special software provided by the video card manufacturer and loaded into the process of launching the operating system. Video driver functions as an interface between a system that is running in her video application and Cartan. As well as video-BIOS, video driver software organizes and supervises the work of all parts of the video through a special Cartan control registers, access to which is via an appropriate bus.

Characteristics of video cards
memory bus width is measured in bits - the number of bits transmitted per cycle. An important parameter in the performance of the card.
memory capacity, measured in megabytes - the amount of its own memory card. Greater volume does not always mean better performance.
Video card, integrated into the chipset on the motherboard or are part of the CPU, do not usually have their own video and use for their own use of the computer's memory (UMA - Unified Memory Access).
core and memory - are measured in megahertz, the greater the faster video card will handle the information.
texture and pixel fill rate, measured in millions of pixels per second, shows the amount of information displayed at a time.
conclusions of cards - video Cartan MDA, Hercules, CGA and EGA equipped with 9-pin type D-Sub. Occasionally, also attended the coaxial connector Composite Video, allowing to bring black and white image on a television receiver or monitor is equipped with low-video input. Cartan VGA video and later tended to have only one slot VGA (15 pin D-Sub). Occasionally, early versions of VGA-adapter plug also had the previous generation (9-pin) for compatibility with old monitors. The choice of working out wondering jumpers video Cartan. Currently, board equipped with DVI or HDMI, Display Port or in quantity from one to three. Some of the latest generation of ATi graphics cards are equipped with six video outputs. DVI and HDMI ports are evolutionary stages of development of a standard video signal, so the connection of devices with these types of port adapters can be used. DVI port comes in two varieties. DVI-I also includes the analog signals, allowing to connect the monitor via an adapter to plug D-SUB. DVI-D does not allow it. Display Port allows you to connect up to four devices, including speakers, USB-hubs, and other input and output devices. The video card is also possible to place the composite and S-Video video outputs and video inputs (denoted as ViVo).

Generation of 3D-accelerators
The earliest boosters used Glide - API for three-dimensional graphics, developed by 3dfx Interactive Video on the basis of its own GPU Voodoo Graphics.
Next generation of accelerators in the video card can be considered under the version of DirectX, which they support.There are the following generation:
DirectX 7 - card does not support shaders, all images are drawn texturing;
DirectX 8 - support for Pixel Shader 1.0, 1.1 and 1.2, DX 8.1, and even version 1.4, support for Vertex Shader 1.0;
DirectX 9 - Pixel Shader 2.0, 2.0a and 2.0b, 3.0;
DirectX 10 - support for unified shaders version 4.0;
DirectX 10.1 - Support Unified Shader 4.1;
DirectX 11 - support for unified shaders version 5.0.
Also, generation of accelerators in the video card can be considered under the version of OpenGL, they support:
OpenGL 1.0
OpenGL 1.2
OpenGL 1.4
OpenGL 2.0
OpenGL 2.1
OpenGL 3.0
OpenGL 3.1
OpenGL 3.2
OpenGL 4.0
OpenGL 4.1

Video Cards Interface 
The first obstacle to higher performance video systems - is a data interface that is connected to the video card. No matter how fast the processor video Cartan, most of his opportunities remain untapped unless provided by the appropriate channels of information exchange between them, the CPU, memory, and additional video devices. The main channel for data transmission is, of course, interface bus of the motherboard, through which is provided to communicate with the CPU and RAM. The earliest bus used in the IBM PC was the XT-Bus, she had a bit 8 data bits and 20 bits of address and worked at a frequency of 4.77 MHz. Next came Bus ISA (Industry Standart Architecture - the industry standard architecture), respectively, it had a bit 16/24 bit and worked at a frequency of 8 MHz. Peak capacity was slightly greater than 5.5 MiB / sec. This is more than enough to display text and graphics shestnadtsatitsvetnoy games. Jerk was the emergence of further bus MCA (Micro Channel Architecture) in a new series of computers PS / 2 firm IBM. She already had a bit 32/32 bit and peak bandwidths up to 40 MiB / sec. But the fact that MCI is a closed architecture (owned by IBM), has prompted other manufacturers to look for other ways to increase the capacity of the main channel of access to video Cartan. And now, with the advent of processors series 486, was proposed to connect to peripherals of the processor local bus, was born as a result of VLB (VESA Local Bus - local bus standard VESA). Working on external CPU clock speeds, which ranged from 25 MHz to 50 MHz and having a bit 32-bit bus VLB provides peak bandwidth of about 130 MiB / sec. This was already more than enough for all existing applications, in addition, the possibility of using it, not only for video Cartan, there are three slots to connect and provide backward compatibility with ISA (VLB is just another pin for 116 slot ISA) to guarantee it long enough life and support for many manufacturers of chipsets for motherboards and peripherals, even in spite of the fact that at frequencies of 40 MHz and 50 MHz to provide service even two devices connected to it is problematic due to excessive load on the stages of the CPU (after all, most managers chains came with VLB to the processor directly, without any buffering). Still, given the fact that not only the video card began to demand high-speed information exchange, and the apparent inability to connect to all devices VLB (and need for a middleware solutions are not limited to PC), was developed bus PCI (Periferal Component Interconnect - integration of external components) appeared in the first place, on motherboards for processors Pentium. In terms of performance on a PC all remained the same - when clocked at 33 MHz bus and bit 32/32 bit it provides a peak capacity of 133 MiB / s - as much, and VLB. However, it was more convenient and, eventually, replaced the tire and VLB on motherboards for processors of class 486.
With the advent of Intel Pentium II processor and a serious bid for membership in the PC market high-performance workstations, as well as with the advent of 3D-games with complex graphics, it became clear that the capacity of the PCI, as it existed on the platform of PC (usually at 33 MHz and bit 32 bit), there will not be enough to satisfy the requests of the system. Therefore the firm Intel has decided to make a separate bus for the graphics subsystem, some upgraded bus PCI, provided the resulting new bus, a single memory access with support for some specific requests of the video is called Cartan and AGP (Accelerated Graphics Port - accelerated graphics port). Bit AGP bus is 32 bits, 66 MHz operating frequency. The first version of the interface supports data transfer mode 1x and 2x, second - 4x, third - 8x. In these modes are transmitted per clock cycle, respectively, one, two, four or eight 32-bit words. AGP versions were not always compatible with each other in connection with the use of different supply voltages in different versions. To prevent damage to the equipment used in the key slot. Peak throughput mode 1x - 266 MiB / sec. Issue-based video Cartan PCI and AGP for the moment is negligible, since the AGP bus did not meet modern requirements for the power of new PCs, and, moreover, can not provide the necessary power supply. To solve these problems created by the expansion bus PCIE - PCI Express 1.0, 1.1 and 2.0. This sequence, in contrast to AGP, an interface, its capacity can reach several tens of GB / sec. At the moment, there was almost complete abandonment of the AGP bus in favor of PCI Express. However, it is worth noting that some manufacturers still offer enough advanced in their design video card with PCI or AGP - in many cases it is a fairly simple way to dramatically improve the performance of obsolete PCs in some graphics problems.

Video Memory
In addition to the data bus second bottleneck of any video Cartan - it's capacity (born bandwidth) of the video memory of Cartan.Moreover, the initial problem arose not so much because of the speed of video processing (which is now often face the problem of information "hunger" video controller when it processes the data more quickly than they have time to read / write to / from video memory), but because of the need access to them from the GPU, CPU and RAMDAC'a. The fact is that at high resolutions and high color depth to display the screen on the monitor should read all the data from video and convert an analog signal, which will go on display, many times per second, how many frames per second display shows. Take a one-page display at 1024x768 pixels and a color depth of 24 bits (True Color), it is 2.25 MB. At 75 Hz frame rate need to read this page of video memory Cartan 75 times per second (the read pixels are transferred to the RAMDAC, and it converts the digital data of the color pixel into an analog signal fed to the monitor), and nor linger, or miss a pixel can not be thus, nominally Required memory bandwidth for a given resolution is approximately 170 MB / s, and this is without taking into account the fact that it is necessary and most video controller to read and write data from this memory. To resolve 1600x1200x32 bit at the same frame rate 75 Hz, nominally Required bandwidth is already 550 MB / sec. For comparison, the processor Pentium-2 had a peak speed of memory 528 MB / sec. The problem can be solved in two ways - either to use special types of memory, which allows two devices to simultaneously read from it, or put a very fast memory. The types of memory and will be discussed below.
FPM DRAM (Fast Page Mode Dynamic RAM - dynamic random-access memory with fast access paged) - the main type of memory, identical to those used in residential motherboards.Uses asynchronous access, in which control signals are not tied rigidly to the system clock frequency. Was actively used until about 1996
VRAM (Video RAM - Video RAM) - the so-called dual-port DRAM. This type of memory provides access to data from the two devices at once, ie, it is possible to simultaneously write data to any memory cell, and simultaneously read data from some neighboring cells. In this way allows you to combine the time-image output to the screen and processing it into memory, which reduces delays and increases access speed. RAMDAC that is free to display on the screen over and over again on-screen buffer, not at all disturbing video processor to perform any data manipulation. But it's the same speed DRAM and it is not too high.
WRAM (Window RAM) - version VRAM, an increase of about 25% capacity and the support of some commonly used functions, such as rendering fonts, moving image blocks, etc. It is used almost exclusively on the accelerator firm Matrox and Number Nine, because it requires special methods for accessing and processing data. The presence of only one manufacturer of this type of memory (Samsung) greatly reduced the possibility of its use. Video Cartan constructed using this type of memory do not tend to drop in performance when you install the high resolution and screen refresh rates, single port on the same memory in such cases, the RAMDAC takes more time access to bus memory and video performance can be very Cartan fall.
EDO DRAM (Extended Data Out DRAM - dynamic RAM with extended data retention time out) - the type of memory elements pipelining, allowing multiple units to accelerate the exchange of data from video memory is approximately 25%.
SDRAM (Synchronous Dynamic RAM - Synchronous dynamic random-access memory) replaces the EDO DRAM and other one-port asynchronous memory types. Once produced the first reading from the memory or the first entry in the memory, the next read or write operations take place with zero latency. This ensures the fastest possible read and write data.
DDR SDRAM (Double Data Rate) - SDRAM option to transfer data from two cuts of the signal, we obtain as a result of doubling the speed. Further development is still in the form of another seal of packets in one cycle bus - DDR2 SDRAM (GDDR2), DDR3 SDRAM (GDDR3), etc.
SGRAM (Synchronous Graphics RAM - synchronous graphics RAM) DRAM option with simultaneous access. In principle, the work is identical SGRAM SDRAM, but additionally supported by some more specific functions, such as block and mask entries.Unlike VRAM and WRAM, SGRAM is a single port, but can open two memory pages as one to emulate dual-port memory of other types.
MDRAM (Multibank DRAM - mnogobankovoe RAM) - version DRAM, developed by MoSys, organized as a set of independent banks amount to 32 Kib each operating in pipelined mode.
RDRAM (RAMBus DRAM) - memory that uses a special data channel (Rambus Channel), which is a data bus width of one byte. This channel can transmit information in a very large flows, the highest data rate for a single channel at the moment is 1600 MiB / s (800 MHz, data is transmitted on both slices of the pulse). One such channel can be connected to several memory chips. This memory controller operates with a single channel Rambus, single-chip logic can accommodate four such controller, so theoretically can support up to four of these channels, providing a maximum bandwidth of 6.4 GB / sec. Less of this memory - you need to read large blocks of information, otherwise its performance drops dramatically.

The main producers of Video Cards
AMD
nVidia
Intel
Specialty:
Matrox
3D Labs
Other vendors:
Number Nine Visual Technologies (acquired by S3 Graphics)
3dfx (acquired by NVidia)
Alliance Logic
Ark Logic
ATI (acquired by AMD in 2006)
ASPEED Technology Inc.
Bitboys Oy (acquired ATI in 2006)
Chips & Technologies (acquired by Intel in 1997)
Cirrus Logic (left the market in 1996)
Hualon Microelectronics Corporation
Macronix
NEC
NeoMagic Corporation
Oak Technology (left the market in the early 90s of the twentieth century)
Real3D (SP, Intel and Lockheed Martin, the actual developer i740)
Realtek
Rendition
S3 Graphics (graphic unit acquired in 2000, VIA)
SiS
Trident Microsystems (in 2003, a graphic department purchased XGI Technology Inc.)
Tseng Labs (acquired by ATI in 1997)
VIA
Western Digital (left the market in 1989)
XGI Technology Inc. (Acquired by ATI in 2006)