DirectX10 explanation

I was flipping over a few pages of graphics cards reviews (since I've missed so many of them!) and suddenly I found an article, previewing DirectX10.

The article was good, especially when it explained the geometry shader in a plain English. I guess DirectX10 promises a very bright future in graphics indeed.

"The traditional vertex shader takes in a single vertex, a single point, and it has to output a single vertex. It's impossible for the vertex shader to create or destroy triangles because it always has to output a vertex for each one it takes in. The geometry shader lets the game operate on entire geometry primitives, lines, triangles, and points as well as neighboring adjacent primitives. The geometry shader can also create new primitives, add new triangles, before sending them further down the pipe to the rasterizer and pixel shader.

Even better, the geometry shader can also output results directly into memory, allowing data to go right back into the graphics pipeline without going out to the CPU for setup. The changes will let the GPU accelerate particle effects like smoke and explosions, which are usually done on the CPU. Games can also use the geometry shader in combination with texture arrays to accelerate effects like cube mapping. Games use cube maps to make an object reflect the world around it. To make an object like a stainless steel teapot reflect its surroundings, you have to determine the teapot's surroundings and map it onto the teapot. Normally, this can take six passes, but a game can use the geometry shader and a render target array to create a cube map in one pass."

Head over to TweaksRUs for the rest of the article.