I have a global illumination renderer which spits out about 170MB worth of textures (2MB compressed JPEGs) for my current level. Naturally, I don''t have the video RAM to cache all these, so I''m trying to implement a simple LOD system to size down the images to 1/2. 1/4, 1/8 their size depending on distance from the camera before rendering them. Here is my problem: - Every time the camera moves, some textures need to be recomputed. - I don''t want to keep all 170MB in RAM, so I have to decompress the JPEGs and then filter them down, which takes a long time and drops the frame rate. - I can''t pre-filter the images, because storing mip-maps would take 4x170=680MB, which I don''t have. Any ideas? Tom

Anybody?