<--back--      home     

What is going to happen next?

Render to texture will allow us to bind an off screen renderer as a texture, which will avoid the expensive copy operation needed for many multi-pass rendering techniques. (Now if only we could modify the position of a fragment...).

I am told that faster read backs are on their way. A big bonus for parallel rendering.

I speculate that we will see hardware features which target point based surface models. We saw several papers this year on the topic.

The number of texture units per polygon will increase with each generation until we just don't know what to do with them all.

I expect the number of vertex operations to also increase.

We should be able to create and destroy vertices and polygons soon.

I hope and pray that we get more bits of precision in the frame buffer, and soon!

Stanford is already hard at work on a higher level shading/rendering language.

The nForce architecture is an interesting development, which may hold the key, at least near term, to fixing some of those bus issues I mentioned earlier.

This is some cool hardware.