PCI Express

Since its launch in 1997 with the Intel 440LX chipset AGP (Advanced Graphics Port) has served a substantial role in PC graphics performance taking over from the much limited shared PCI bus architecture. It was first thought that memory would be expensive and thus future graphics cards would also utilize system memory, later this theory turned out mostly invalid but moving textures and other data to and from the graphics sub-system still required a fast independent bus. In its latest revision AGP 3.0 or more widely known within the IT community as AGP 8x provides a direct maximum bandwidth to the graphics card of 2.1GB/sec. Taking over this role PCI Express running at 2.5GHz has a peak data throughput determined by the number of channels employed into a given socket, in terms of add-on cards this is currently x1 (a single channel) but for graphics cards it's a full x16 (16 independent lines). In practice this means that Galaxy's 6600GT and other 164 pin PCI-E x16 products have at their disposal no less than 8GB/sec (bi-directional of pure bandwidth, 2x 4GB/sec), an immense technological leap. Such advancement will continue to push graphics sub-system I/O performance further especially as game titles continue in gaining increasing complexity pushing yet more data down through to the graphics card. The most exciting aspect of PCI Express is SLI as mentioned earlier, for those truly geek orientated individuals craving no less than two graphics cards in their system running in parallel it is this technology that most certainly brings smiles to their faces.

High-end technology

As already briefly stated Galaxy's 6600GT is a fully compliant DirectX9C graphics card, in simple terms this means that any eye candy offered through this broadly used API usually in games will most definitely be shown on screen providing the given game makes use of it. That however is only the beginning as Nvidia and equally ATI Technologies promote their GPU's through adding further competing technologies that at many times sound compelling, occasionally bring real-world large benefits but most of all have the potential to confuse the consumer beyond belief. As a result of this marketing misfortune we will attempt to put some words of wisdom into making the most interesting techno babble somewhat more comprehendible.

CineFX 3.0

This technology or rather subset of technologies is Nvidia's new 3D engine aiming to take reality to the next level. The third generation housing Shader Model 3.0 includes improvements made to both the Vertex and Pixel shader Model. From a consumer point of view this allows developers to write more complex shader routines but also maintain or even surpass performance efficiency over Shader Model 2.0. CineFX 3.0 fully confirms to Microsoft's DirectX 9C API standard and as such allows for amongst other improvements a far greater or in practical terms freedom approach to the number of maximum instructions possible in pixel and vertex shading paradigms. In comparison to Shader Model 2.0 the maximum instruction count for pixel and vertex operations was 96 and 256 respectively, within Shader Model 3.0 this is now 65535+ and 65535. In real world and non theory terms such advancement opens the window to enabling virtually limitless and thus very realistic lightning and texturing scenarios, something game developers will be taking advantage of well into 2005, most game titles at present still employ Shader Model 2.0. By also enforcing 32bit floating point precision textures themselves become more vibrant and yield greater clarity, a feature absent from SM 2.0. In order to not hinder processing efficiency by increasing scene complexity SM 3.0 introduces flow control which itself introduces refinement in how developers program shading into their products. By introducing looping/branching and subroutine call/return functions less code is necessary to perform the same visual effects, it is also as a result more condensed and when combined can yield significant saving of GPU resources but also memory bandwidth. This can be for example ilustrated by instancing, a new feature allowing programmers to create multiple versions of the same object without having to reconfigure each sample individually - only their differences in behavior must be modified hence the reduction in otherwise unnecessary overhead. Of course technical details such as these often confuse the consumer and mystify their view of what technology or feature is more commendable, CineFX 3.0 thus Shader Model 3.0 as it stands is undeniably a forward thinking and otherwise future friendly technology to consider when purchasing a graphics card. Currently only Nvidia offers SM 3.0 support in their Geforce 6 series GPU's but ATI it set to follow during 2005.