Just something i've been wondering about for a while now. Each pixel drawn is related to the surrounding pixels in some way (e.g. Anti-aliasing, blooms, overdraw, motionblur, etc.). When you chop a scene in two, the part of the scene close to the edge would still need to render some pieces of missing sections in the other half in order to render its own scene properly. So rather than the job actually being split 50%/50%, it'll actually be something like 55%/55% right?
That's with just one edge. If you were to do supertiling and do the scene in a checkerboard pattern, wouldn't both cards just practically be rendering the same scene twice since they're both filling in the missing pieces in order to do certain effects?
A 20 pixel high overlap may not be much for scenes split in two, but if the entire scene is covered with overlaps wouldn't that cancel out the benefit of having two cards?
I've been googling and they don't actually explain how it's supposed to be faster, they just define what supertiling is.
That's with just one edge. If you were to do supertiling and do the scene in a checkerboard pattern, wouldn't both cards just practically be rendering the same scene twice since they're both filling in the missing pieces in order to do certain effects?
A 20 pixel high overlap may not be much for scenes split in two, but if the entire scene is covered with overlaps wouldn't that cancel out the benefit of having two cards?
I've been googling and they don't actually explain how it's supposed to be faster, they just define what supertiling is.