All the current consoles have spec-sheets with claims of texel fillrates as high, and in xbox case, double the pixel fillrate. I know fillrate is dependent on a number of factors, bandwidth, latency and cache arrangement to mention a few, but the average console game has, what, three to six megabytes of texture bandwidth available per frame? That is enough for at least six 512 x 512 16 bit textures and if 4 bit texture compression is used - like almost all consolegames does - twenty-four textures of that size.
There is of course overdraw to subtract from that figure, but as almost all games are doing some tiling of the textures, and single large textures, most of the time, use only the lower mip levels. That should more that make up for the overdraw.
My question is then why, with all this available texture mem/bandwidth, don’t we have perfect 1:1 texel to pixel coverage yet? It seems like the technology is there to do it.
And now for the question - sort of related to the first - about S3TC. If you look closely at texels on some large texturemaps in many games, it’s quite clear that S3TC couldn’t have been used, as the resolution of the maps is too low, and the number of colours to high, to make it possible to get a satisfactory result with S3TC (look at for example the some of the ground textures in Sudeki, where in some cases it is hard to find a 4x4 square which uses only two colours with two hues (the requirements of S3TC)).
My question know is, how much is 4 bit S3TC - or any of the DXTC variants - really used in actual games? To me, only by looking at the graphics, it seems as though the preferred “compression†is multipass texturing.
There is of course overdraw to subtract from that figure, but as almost all games are doing some tiling of the textures, and single large textures, most of the time, use only the lower mip levels. That should more that make up for the overdraw.
My question is then why, with all this available texture mem/bandwidth, don’t we have perfect 1:1 texel to pixel coverage yet? It seems like the technology is there to do it.
And now for the question - sort of related to the first - about S3TC. If you look closely at texels on some large texturemaps in many games, it’s quite clear that S3TC couldn’t have been used, as the resolution of the maps is too low, and the number of colours to high, to make it possible to get a satisfactory result with S3TC (look at for example the some of the ground textures in Sudeki, where in some cases it is hard to find a 4x4 square which uses only two colours with two hues (the requirements of S3TC)).
My question know is, how much is 4 bit S3TC - or any of the DXTC variants - really used in actual games? To me, only by looking at the graphics, it seems as though the preferred “compression†is multipass texturing.
