Group test graphics cards.

Byline: Emil Larsen

As always, there are huge changes in the way the latest generation of graphics cards operate, the biggest being the shift to using unified shaders. In the past, graphics programmers could perform only a very specific array of functions because each graphics card pipeline could only accept certain commands. Then programmable pipelines were introduced, which meant developers could write their own mathematical equations to create customised visual effects. This can give each game a unique look and, through experimentation, graphics have improved no end.

Programmable pipelines come in two main types - vertex shaders and pixel shaders. A vertex shader manipulates the corners of triangles, which all 3D scenes are made up of. Vertex shaders breathe life into characters and their environments by bending and stretching objects. For example, it can give rise to facial animation where wrinkles appear when a character frowns.

Pixel shaders, meanwhile, apply effects to scenes afterwards on a per pixel basis. Each pixel (there can be millions) gets a texture, is lit up and has other effects applied to it, such as advanced reflections. In the past, if you were playing a computer game level that is vertex heavy, the pixel pipeline's transistors simply wouldn't be used. Similarly, if you were playing a level that's pixel heavy, a lot of the vertex pipelines go to waste.

On their latest graphics cards, Nvidia and ATI have created unified shaders where each pipeline can perform pixel or vertex operations to get the most out of the card. Current unified shaders can also perform physics operations - for example, complex collisions that might otherwise be calculated on the CPU. Geometry shaders are part of the DirectX 10 specification and are considered by many to be the most important step forward in recent years. Vertex shaders are limited to generating one output vertex from each input it receives, whereas geometry shaders can create several vertices in a single go, which means more complex animation for less processing power.

Unified shaders, which ATI and Nvidia call stream processors, are present on all of the latter's Geforce 8 series cards up and, similarly, all of AMD's Radeon HD models (with the exception of the inappropriately named Radeon HD 2300, which is a Radeon X1300 with extra video decoding functionality). The number of shaders you have reflects the processing power available, rather than the clock speed.

It's important to understand both ATI's and Nvidia's stream processors are different and not directly ...