Shade maps = awesome , whats the catch?

[pg1]

Just started using shade maps for stereo and they seem brilliant.
Can anyone explain why the shade version of a render besides being faster also seems to be a lot less noisy?
Also there is an obvious difference between a render done with 1 reuse threshold (slow, less shade map, more "correct") compared to a render done with 100 re use threshold (faster, more shade map, less "correct")
I dont really understand why the slower more "correct" render is better? The edge of an object doesn't move in either so does the difference between the renders effect how 3D it looks?

[instinct]

Mind you, guessing based on the info shared on the shademap. I might be wrong there.

From what i understood shademaps work similar to irrmaps to some extend. So basically the first view is beeing baked to a 3d point cloud that is re-used for the second and only
missing parts are rendered. That also explains the second beeing smoother, because it is interpolating the map. The lower the res, the smoother, the less detail.
Reusing the shaded result is obviously way faster than calculating it, but it is also a bit wrong, because a reflection would obviously be slightly different between the views.
As eyes are rather close this is something you usually can live with tho.

Regards,
Thorsten

[pg1]

Thanks instinct,
that confirms what I thought was happening. Got any rule of thumb for the reuse threshold, from quick tests I found that it differs from scene to scene.
I suppose my real question is would it be a good idea to leave the reuse threshold to 100 all the time so it uses all the shade map?

[instinct]

I have no real world experience yet that would allow such a bold statement hehe. I guess it is similar with anything that bakes+interpolates a matter of many factors
in a scene (also depending if you also use it for moblur and have moving objects e.g.?). I'd say start with full shademap and pull back if needed? That might lead to
rerendering tho heh.

Regards,
Thorsten