|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
Gigabyte X800 Pro Review :: Features
The X800 Pro is a heavily modified and improved R300 (RADEON 9700 Pro et al). In a sense if you take the 9700 Pro and increase the pixel pipelines to 12 and the clock speed to the clock speed of the X800 Pro you would get almost a X800 Pro. I'm over simplifying here but that's the general idea. The X800 Pro has improved support for the features of DirectX 9.0 compared to the R3xx series. First up is Pixel Shader 2.0b. The 9800XT and other variants of the R3xx series was capable of 160 pixel shader operations in a single pass with multiples of that available with multi-passing. With the X800 series, ATI increased their maximum pixel shader instructions in a pass to 1536 consisting of 512 texture instructions, 512. The instruction set of the X800 Pro is exactly the same as the 9800XT or other R3xx series cards with the addition of a facing register for two sided lighting. ATI felt perhaps rightly so, that the amount of good DX9 games that would require Pixel Shader 3.0 not to number very many. One could say that there aren't very many good DirectX 9 games requiring Pixel Shader 2.0 either. Vertex Shader 2.0 is the standard that ATI chose to follow with the R4xx series (which encompasses the X800 Pro). The X800 has 6 parallel Vertex Shader Engines, which is an increase over the 4 that the R3xx series has. With the increase in clock speed and other improvements to the vertex shaders on the X800 Pro the X800 Pro is capable of 700 million vertices a second and 65,028 vertex shader instructions per pass with loops and branches. So what is a normal map anyway? When lighting a 3d object in a game it's important to know what direction the object is facing from the light source in order to correctly light the object. After all you don't want to light everything. A normal is the direction that a point on the 3D object is facing. You draw a perpendicular line from the point on the object. Now draw a line from the point to the light source. This gives you a light vector (line). The angle between the normal and the vector is how you tell how much lighting is needed. Now a normal map is a special texture map that holds information on how that object should be lighted. Normal Maps have several excellent uses. First it can add more detail to an object by adding or subtracting a bump map from the surface of an object. Simply create a bump map, convert the bump map to a normal map and add it to the surface of an object and voila you got extra detail. The second use is to give the illusion of a higher resolution model by using the same normal model on a low-res model as on the high-res model. This is what's done in the Double Cross demo. 3DC is ATI's proprietary normal map compression algorithm. It is similar in concept S3TC in many ways. It uses normal maps to give the sense of more detail in programs that use it. The compression ratio is 4:1 like the compression on the DXT5 standard, but the quality of the compression should be a lot better than DXT5. You might hear about NVIDIA's GEFORCE 6800 and it's variants the GT and Ultra and something called Shader Model 3.0. The X800 Pro does not support SM 3.0, but with the scarcity of SM 2.0 games (Pixel/Vertex Shader 2.0), I don't see the imminent need for it at the moment. I'm sure the 6xxx series has its own set of positive attributes besides SM3.0, but since I don't yet have time with the card I'll concentrate on this card.
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
