what you mean aermartin?.. that with ati the image tiling works better than nvidia?
I noticed that there is a peak for efficiency. Mostly, i found that a 32x32 tilling is a good one for the stuff i usually do.
Yet, yesterday, i test my ram increase by rendering the Blender Benchmark picture. In that case the fastest is with 16x16. Going up towards 128x128 has shown a slow down of rendering.
I haven’t used C4D sins quite a time, but wasn’t there some tilling automation at the end phase of renders? And will that sort of thing still be up to date when GPU rendering will get generalized?
nah its just that you generally get more stream cores in your ati card for the same price as nvidias.
besides I dont think you can send a whole tile, as e kernel to be executed on the GPU , but maybe.
I cant wait until a dev w/ ocl checks out the render, if raytrace can be improved, or occlussions passes.
Almux,
I had the same experience with setting the tiles.
I assume at one point it gets too much work to manage the amount of tiles
so it results in slowing down Blender.
yes i also noticed a slowdown increasing tiling too much… i think the best number change from situation to situation.
however, as i asked before, as anyone ever seen a terragen 2 rendering? in the process you can see parts of image as big as dots popping out… but still i don’t know the method used in the program.
ok…since we are speaking of hardware performances i would like to ask you here another question, not to open another thread.
in the 3d view, is the gpu that is working to build the scene, right? if it is like this, why my 250gts is working in 2D frequences when i work in blender viewport? shouldn’t it work in 3d performance mode, and use the higher frequences?
my system start to be not so fluid around 1,5 millions poligons with 2.53, and i read of people working with over 20 millions with similar video cards… how’s possibile?
thanks again for the help!
GLSL in viewport settings?
Guess it is worth pulling up tilling when there are transparencies and mirroring spotting up all over, but placed here and there while included in a rather simple textured scene for the rest.
This way, the high calculation parts get more isolated and the processors can concentrate on those.
mmmm… to be honest i don’t know. i do i check in 2.53?