You could be forgiven for misunderstanding where the images on your screen come from. Monitors visualise the projection of pixels translated by the Graphics Processing Unit (GPU), itself translating the stream of Binary Code flowing from the Central Processing Unit (CPU), turning the entire process into the pictures on our screens. They are found in some format across nearly all varieties of computers ranging from mobile phones to game consoles to industrial processing units. They enable visual interaction with the terminal in question as well as providing enhanced graphical offerings for film, design and gaming.


Consisting of a specialised electronic circuit designed to rapidly manipulate images, Graphical Processing Units are found in a vast spectrum of computer based technologies. Whilst many early computers relied on their CPU’s for image manipulation companies such as Intel and IBM realised that additional processing power would be needed to facilitate advanced computing functions. In 1999, GPU mainstay NVIDIA entered the market with the GeForce 256, promising ‘rendering engines capable of processing 10 million polygons per second.’ Their continual rival, ATI Technologies countered with their own version in 2002, the Radeon 9700. Since then, it has been a continual GPU arms race between the two companies for the title of most powerful processing unit.

Having the most powerful GPU does not always relate to the best performance. As with most computer related components the GPU must be selected in relation to other existing articles within the computer habitat such as the CPU, the RAM, Motherboard and Monitor. The function of the GPU is to visualise our actions within the habitat turning our key taps and mouse swipes into translatable pixels on our monitor.

The GPU essentially functions as a massive graphical calculator located in the heart of the computer. A standard GPU houses its own RAM, a dedicated processor and utilises its own BIOS which maintains the GPU settings and functionality. The GPU can be used to process 3D images at an extremely high rate, first drawing a line image then rasterising the outline with pixels to generate the final visual output. Throughout image intensive modern games a GPU will work under extreme duress, perhaps drawing and redrawing images over 30 times a second. We are increasingly noting an increase in the number of transistors present on GPU boards and along with an increase in the RAM available to the unit, graphics processing is consistently increasing in power. However, this process generates massive amounts of heat so most GPUs will come with a manufacturer issue fan or heat sink. After-market (post-purchase) heat maintenance options are available ranging from simple fans to water and chemical exhaust cooling systems for extremely powerful computer setups.

GPUs need somewhere to store the reams of images they generate so capably. The inbuilt RAM provides the housing for this, buffering frames until they are ready to be used as well as storing data concerning the current pixel usage, the colours in use and their location on the screen. There are also additional functions that GPU RAM can perform such as Z-Buffering (managing coordinates and locations of images, graphics, textures, shaders and vertices), or utilising the RAMDAC function (Random Access Memory Digital-to-Analogue Convertor) which converts and maintains signal differences between digital and analogue computer components.


Graphics Card Interfaces

Before the GPU can even attempt to perform any geometrical calculations, or provide you with some quality pictures of cats it must be connected to the Motherboard, where it can communicate through one of three common interfaces:

  • PCI: Peripheral Component Interconnect
  • AGP: Advanced Graphics Port
  • PCIe: Peripheral Component Interconnect Express

Of the three, PCIe is fast becoming the industry standard due to its increasing information transfer rates and its ability to provide links between two separate GPU’s, further enhancing the mathematical output and buffering capabilities of the computer system. It is common to find most modern GPU’s with two separate monitor connections and many now incorporate a third if powerful enough, or contain combinations of the three:

  • VGA: Support for older CRT monitors as well as some connectivity with LCD screens. Offers lowest resolution options.
  • DVI: Provides digital support for modern screens as well as some connectivity with older CRT monitors.
  • HDMI: Provides High Definition resolution options for digital monitors.

Many newer GPUs require a larger amount of power to work and as such will have a separate 4pin connection linked directly to the PSU (Power Supply Unit). For instance, the graphics card in my home computer is a GeForce GTX 660 Ti. GeForce is the make (NVIDIA), GTX 660 the model number and ‘Ti’ identifies the specific card. This card has 2GB RAM and requires a Minimum 450W power supply, connects via PCI Express 3.0, has 2xDVI, 1xHDMI and 1xVGA connections supporting a Maximum Digital Resolution of 4096x2160. Each card should be considered for its merits when compared to the computers main functions, be that video editing, gaming, word processing or product design, different cards will suit different needs.


More Graphics Card Terminology

Some additional GPU related terminology you may encounter when choosing your video card:

  • OpenGL/DirectX: These are Application Programming Interfaces (API) and help hardware and software communicate with specific aspects of the GPU during intense andc. OpenGL is founded and maintained by the Open Source community; DirectX is owned and maintained by Microsoft.
  • Anisotropic Filtering: renders crisper images
  • Full Scene Anti-Aliasing: the edges of 3D objects appear smoother
  • Shader: An aspect of the GPU that describes traits of the pixels. Enables the GPU to render images with the correct levels of colour and light to appear natural. Comes in three main types:
    • Vertex: Transforms a 3D image positing to a screen coordinates for 2D monitor.
    • Geometry: Relatively new to the API’s (see above), these generate new graphics primitives when needed (e.g. basic lines, points, triangles) for the starting points of more complex 3D images before being rasterised.
    • Pixel: Compute the relative depth, frequency and tone of each fragment of colour available to a pixel.
  • Post Processor: Aspect of 3D rendering. Image is first buffered into GPU RAM, where pixel shaders and post processing techniques to complete image:
    • HD Range Render: Dynamic range lighting for high end video processing
    • Bloom: the ‘glow’ effect
    • Motion Blur
    • Depth of Field

More Stuff

All About Things

How To