A computer requires a translator to take binary data (the bits that represent colour and so forth) and convert it into a picture. This usually takes place on a graphics card. 3D images are created by first sketching a wire frame from straight lines before the colours, textures and lighting are added. For fast-paced modern games, this process may be done up to sixty times a second.
This requires significant processing power and if left to the motherboard, with it’s myriad other demands could easily overwhelm it. Hence the graphics cards, designed to take over this process has its own processor and RAM. The GPU, also known as the Graphics Processing Unit, is it’s equivalent of the computer’s CPU. It is designed specifically for performing mathematical and geometric calculations for rendering and some even have more transistors than an average CPU.
As images are rendered, they have to be stored before being display, hence the use of RAM. Typical video RAM operate at high speeds are dual ported. This means the system is allowed to both read and write to it at the same time. This is then connected to a DAC or digital to analog converter. This device converts the images into an analog signal that the monitor uses to display images.
A good indicator of a Graphics card performance is its frame rate. This is measured in frames per seconds or FPS. The components of this frame rate include triangles or vertices per second, the speed at which polygons are calculated to define 3D images and pixel fill rate, which is the number of pixels it can process per second, also the rate at which it rasterize images.
Two of the top producers of Graphics cards today are nVida and ATI who apply processes to enchance GPU performance. These include Full Scene anti aliasing (FSAA) which gives 3D images smoother edges and Anisotropic Filtering (AF) which make images look crisper. Other specific techniques are also used to aid the GPU in application of colours, texture and shading.
Source: http://computer.howstuffworks.com
Just to add, graphic cards can only take some mathematical functions. What it can do, is do hundreds of time of a similar calculation and pass it back to the processor("Core 2 duo") who will then output the information.Most graphic cards comes with little or no cache. Hence, the ability to manipulate multiple processes are limited. But due to the abundance of cores(nvidia gt9800 and above) hae more than 200 cores, which enable it to process up to 200 times faster then what your processor can do! This is also why, the next generation of computing(coded CUDA) would seemingly be an ideal choice for rapid and efficient engine!
ReplyDelete