- Also called:
- graphics card
- Related Topics:
- integrated circuit
- graphics processing unit
video card
What does a video card do?
Can a PC run without a video card?
How have GPUs evolved over time?
video card, circuit board that processes graphical data from a computer and generates a video signal to be sent to a computer display such as a monitor. As opposed to integrated graphics systems, in which the graphics processing unit (GPU) is built into a computerās central processing unit (CPU) or motherboard, video cards are discrete circuits slotted into or soldered onto motherboards.
Components and operation
A discrete video card typically contains a graphics processing unit (GPU), memory chips, display connectors, a motherboard interface, power connectors, and a voltage regulator module (VRM). It also usually has a heat sink, one or more fans, or a liquid-cooling system. Older cards may also contain random-access memory digital-to-analog converters (RAMDACs) meant to convert digital image data (data in binary) into analog display (continuous information) for older display devices, such as cathode-ray tube (CRT) monitors.
A video cardās processing functions are centered on its GPU, a processing chip with multiple cores for calculations. In conventional graphics, a computerās software sends instructions describing a scene to the GPU, which performs the calculations needed to transform three-dimensional models, apply textures and lighting, and determine the color of pixels in the resulting image. Completed images, called frames, are then transmitted to the display. The number of frames produced each second, measured in frames per second (FPS), is an important measure of graphics performance for applications such as 3D modeling or gaming. The GPU performs several calculations to render geometric elements such as pixels, vertices, and planes for video rendering, and typically contains hundreds or thousands of cores (individual processing components) to execute them.
Chips soldered on the video card host its firmware, or Basic Input/Output System (BIOS), and provide VRAM. VRAM is important to expand the processing capabilities of a graphics card as it stores data relevant to graphical rendering, such as image buffers, textures, and other graphical information, in the short term so the GPU can fetch it quickly for faster calculations. The more VRAM a card has, the more graphical data it can store and, if the GPU has a sufficient number of cores, this can speed up graphical rendering considerably.
Most discrete cards connect to a computerās motherboard through a set of connector pins arranged in a wedge along one side of the video card. This provides an interface between the card and the computerās motherboard. It is usually a Peripheral Component Interconnect Express (PCIe) interface, and it provides display connections, such as Video Graphics Array (VGA) for older systems or High-Definition Multimedia Interface (HDMI) for newer ones.
Development
During the 1980s and 1990s, graphics hardware became increasingly capable of rendering two-dimensional and then three-dimensional operations. IBM was one of the first companies to develop the foundational video card, launching the monochrome display adapter (MDA) and color graphics adapter (CGA) in 1981. These cards did not feature GPUs. Instead, they used simple adapters that converted binary data into simple displays. NVIDIA released GeForce 256 in 1999 and marketed it as the first GPU, though the product was essentially a video card using an NV10 silicon chip as a GPU. It combined transformation, lighting, and rendering functions on a single processor and helped establish the modern programmable graphics processor.
As graphics processors became more programmable, their architecture, with its multiple cores working in parallel, proved useful for calculations unrelated to displaying graphics. Technologies such as NVIDIAās CUDA allow software developers to use GPUs for general-purpose computation. GPUs consequently became important not only for computer games, animation, video editing, and three-dimensional design but also for machine learning, scientific computing, engineering simulation, data analysis, and high-performance computing.

" height="20" alt="HEW Logo">