doreamon wrote:I'm getting drastically different renders with bidirectional pathtracing vs bidirect mlt.
Thios should NOT be the case, both render modes should end up in the same result!
maybe this is a rainhard tonemapping issue, since it adapts tonemapping to the scene, some fireflies can trick the tonemapper to adjust diferently!
Doublecheck if this happens with linear tonemapping too, if it does please send scene in pigs format to the devs so they can take a look!
doreamon wrote:I wanted to speed up render and not use pathtracing(I know its more accurate)
No, all rendermodes are unbiased and therefor equal in accuracy (if rendered loooong enough!).
But the different rendermodes have different strengths and weaknesses. For interior using BiDir Pathtracing is usually the best choice in the very most situations. MLT is great in doing caustics for glass, godrays and tackling complicated scenes with very small light sources (raise the MLT setting Large Mutation Prob to 0.85 for a less patchy render noise) but MLT is less effective in the long run, taking longer to kill the rest of the minimal noise in long renderings...
doreamon wrote:Doesn't supersampling just render the scene x=n times? Wouldn't that increase rendertime more?
Isn't faster to render a scene 10000 pixels then scaling down x5 in photoshop.
No, supersampling only cost RAM. Your Photoshop approach would end up in the same result
