The NV-31 GPU is a fascinating chip to say the least. With all the new technology that it uses I am hoping to see games come alive in the near future. The NV-31 is a 0.13 micron GPU with 80 million transistors.  To keep this in perspective, the Intel Pentium 4 Northwood has 55 million transistors, that's 46% more transistors then the Pentium 4, and on the P-4 most of those are used for cache. With the NV-31, the majority is being used on the 3D pipeline. The reason for the high transistor count on the Geforce FX is to comply with the DX9 standard, making the Geforce FX a fully floating point GPU. Going with the 0.13 micron process will also benefit from faster switching transistors that use less power and create less heat.

The Geforce FX GPU was designed to create the same realistic computer scenes that the motion picture studios create but in real time. I don't know about you, but I have been waiting for a GPU to be able to deliver this. One of the enhancements of the Geforce FX is the CineFX engine which is able to provide 128 bit color made up of 32 bit components for red, green, blue, and alpha values. The Vertex and Pixel shaders in the Geforce FX GPU can handle sophisticated pixel shading that are able to create even the most elaborate effects. The 32 bit floating-point pixel precision allows games to display the highest quality images available. The CineFX engine creates 16 and 32 bit floating point formats allowing 32 bit to display breathtaking special affects like ground fog and spherical fog while the 16 bit can offer the optimal balance between image quality and performance. The 16 bit format in fact is what Industrial Light & Magic and Pixar use for there production effects. The greatest part about both these formats is that they can be used on the fly together. For instance, if a developer wants you to be in a courtyard with fog on the ground he would use the 32 bit format for the fog and the 16 bit format for the walls of the courtyard.

Intellisample is a another technology that NVIDIA added to the Gforce FX GPU. This technology is able to display high levels of image quality without hurting frame rates to badly. They do this by using hardware color compression that will compress color information by a 4:1 compression ratio. Due to compression and decompression taking place in real time, there should be no loss in image quality or performance and this allows antialiasing to not hurt performance no mater how high it's set. Fast buffer clears are also a new improvement allowing the buffers to be cleared much faster, saving time and memory bandwidth. Dynamic Gamma Correction and Adaptive Texture Filtering is also another way NVIDIA has improved the image quality with the Geforce FX. Also new with the Geforce FX GPU, is 6XS and 8X antialiasing modes allowing the best image available. The 8X antialiasing will calculate twice the amount of samples as the previous 4X antialiasing.

Microsoft's latest release of DirectX 9 is fully supported by the Geforce FX GPU. In fact, in some cases NVIDIA went beyond the DirectX 9 standards and added even more to the GPU then was needed, giving you support for other effects beyond DirectX 9. Included in the new features of DirectX 9 are Pixel Shader 2.0 and Vertex Shader 2.0. The benefits of these are a more programmable GPU, allowing more lifelike and better performing graphics. With the Pixel Shader 2.0 specification, there's a much broader range of tools the developers can use. For example, procedural shading which is the ability to texture an object mathematically rather then trying to wrap a 2D texture around a 3D object.  Also, there is better Flow Control allowing developers to reuse code to enhance productivity and leave more time for making better effects. Vertex Shader 2.0 makes true programmability for DirectX 9 and also includes new mathematical functions. As an example, developers are able to use longer vertex programs that can have a larger impact on the complexity of a scene like making realistic animated water and allowing the program to decide when additional calculations are unnecessary to improve the image which would improve the overall performance.

The Geforce FX comes standard as AGP 8X which runs at 533.33Mhz, compared to 266.67Mhz that AGP 4X ran at.  This allows the AGP port to move 2.1GB a second.  AGP 4X was only able to move 1.1Gb a second. What does this mean for performance?  Well right now with the games and applications that are on the market it really doesn't mean anything because most software that uses the AGP port was written with AGP 4X in mind.  As programs start getting more graphical and moving more data to the AGP card AGP 8X will almost be a requirement. The AGP 3.0 specifications is what introduced us to the AGP 8X. Some of the other enhancements of AGP 3.0 include the ability to dynamically switch to new scenes in real time, ability to stream data from memory, and backward compatibility to AGP 2.0 systems. As new games and applications come out you will see great improvements in performance by using AGP 3.0