>   According to my experience method 2 gives higher quality results at
> bigger threshold values than method 1.
>   Using a very high number of samples is reasonable only if there are
> *tons* of subpixel-sized details. This is usually rare.

Infinite checkered plane has infinitely small squares.

But they cover such a *tiny* area of the surface...

Similar situation with (say) reflections on a cylindrical shape. Lots of 
small details; you definitely need good AA for it to look right. But I 
don't think I have *ever* been above depth 4...

It would, of course, be trivial to do a back to back compare to see the 
results...

Post a reply to this message