A personal theory about PS3 GPU.

[Urian]

I am going to name it PS3UGP.

The PS3UGP is formed by 2 different cores and an 16MB eDRAM module by Toshiba, the clock speed is 800Mhz.

The first core is the Geometry Engine+Texturing+TMU that renderizes the image in the eDRAM but without effects.

When the first core renderizes the image all the image is divided in pixels, all the data of the resultant image is in the eDRAM.

The second core makes the work of Rasterizer+Full Programmable Color Combiner (Shaders) and remaps the image with all the special effects. After this the Image is put in another place of the eDram and is sent to the TV.

 

[Jawed]

Texturing is too intrinsic to shading for you to be able to pipeline them as separately as you propose.

How would a shader rener to texture and then re-use that texture, or have another shader use that texture?

How would you use textures as the basis for mathematical functions (i.e. as look-up tables, or pre-computed functions) in shader code?

Jawed

 

[version]

ps3 with 2 CELLs and GPU

a cell for geometry other for shaders and physics, and gpu for texture and raycast