Creative Labs3d Blaster GeForce3 Ti500

Author : Wayne Date : 24th September 2001

3DVelocity would like to thank Creative Labs and especially Rosie Tickner of ProdigyPR for their help and courtesy in providing this graphics card for review.

 

Lossless Z Compression :

Every pixel to be rendered by a graphics card has a Z value. This value tells the graphics processor how deep that pixel sits in the final scene, deep meaning how far from the viewers eye point. These values are stored in part of the frame buffer known as the Z buffer. Obviously with moving scenes these values are dynamically changing and traffic to and from the Z buffer can be huge as depth values are constantly read and written for every single pixel. Using the Lightspeed Memory Architecture’ s Z compression/decompression engines, the amount of data is compressed by a factor of 4 meaning the same information can be read and written using only a quarter of the normal bandwidth. This process is carried out in real time and in hardware, and the fact that it is a lossless compression technique means that accuracy is not affected in any way.

Visibility Subsystem: Z-Occlusion Culling :

For every scene a graphics card creates, part of what is calculated will be wasted. By the very nature of a 3D scene, some things will be hidden by other things that are in front of them, and unless steps are taken to correct it, even parts of the final scene that will ultimately be hidden are rendered fully before the graphics processor discards that information on the basis that it won't be seen. Clearly we need a routine that is able to decide what can and can't be seen before the data is processed, and this is where Z-Occlusion Culling comes in. By looking at each pixel early in the processing stage, it is determined whether or not a pixel will be visible in the final scene and if not that pixel is not processed. NVIDIA are a little sketchy on how this is actually achieved, and in all honesty theirs is not the most efficient form of HSR (Hidden Surface removal) around, but in combination with the other bandwidth saving technologies it certainly helps. Another technique employed is “Occlusion Query” which allows a developer to instruct their application to ask the GPU to generate a "bounding box" over a part of a scene to test for visibility in the final rendered scene. This is almost a "forced occlusion" technique, but because it relies on specific calls from the application there's no way of knowing how, when or indeed if