ok, probably going against the tide here, but would a bucket based RT renderer be able to handle higher poly scenes? im thinking specifically about dynamic geometry here, fe. displacement, you have to generate the whole mesh on the cpu and load it onto the gpu in one go. this is inefficient and limits the displacement detail.. if the rendering was tiled, or bucket based, the gpu could render one chunk while the cpu was generating the mesh for the next and so on.
same goes for proxies. at the moment the whole proxy is loaded before the gpu starts, so you get no memory benefit from the proxy system.
same goes for proxies. at the moment the whole proxy is loaded before the gpu starts, so you get no memory benefit from the proxy system.
Comment