[Bubbaloo]

But remember in that test, you are comparing triangles that have varying normals. Comparing 2 triangles that face the same direction is not the same as a box with 6 sides with normals facing 6 different directions. You should test with a sphere or a plane separately with varying amounts of subdivisions. Also I wonder how smoothing affects render times? A sphere with no smoothing compared to one with smoothed faces.

[brodie_geers]

Fair enough, my point was to test the primary objects I've used to create an emitter though (plane, box, sphere). More tests could be done but I'm not totally sure what, say a 2 triangle plane vs. a subdivided 200 triangle plane would necessarily tell us in terms of "real life" rendering options.

-Brodie

[bjorn.syse]

Also I wonder how smoothing affects render times?

Indeed, a very interesting question.

Aswell, I wonder, is SL really a reliable measure of how much the images have cleared in different lighting conditions? Look at the SSS-example, where the rightmost picture has cleared a lot more, even though it's sampling level is 10 compared to the ones where SL is 18 and 19.