Couple performance questions

[romullus]

1. am i right in thinking that if there's noise in reflection and/or translucency render elements, one needs set GI vs AA value higher?
2. is it ok to set LSM to lower than 1 (e.g. 0,5 - 0,2) if direct element cleans very fast?

[Stan_But]

1. Yep. In my tests it is so
2. About LSM. To set LSM low than 1 is no sense.

[romullus]

1. thanks, i was under same impression.
2. why is that? If there is scene with very simple lighting, where direct element cleans extremely fast and most noise comes from complex materials (think of product shots), is there a much sense to shoot 32-64 direct light samples with each pass? I'll try to conduct some tests when i'll have more time to see if it can speed up rendering.

[FrostKiwi]

2. why is that? If there is scene with very simple lighting, where direct elementare cleans extremely fast and most noise comes from complex materials (think of product shots), is there a much sense to shoot 32-64 direct light samples with each pass? I'll try to conduct some tests when i'll have more time to see if it can speed up rendering.

Setting the Light Sample Multiplier this low effectivly kills MIS, as at this point the whole image is being sampled with the same priority and usually noise generating parts of your image like Lights Dont get the attebtion they deserve.
You could (but shouldn't) go into Devel/experimental and turn off MIS completly.
That being said, corona tends to oversample edges on scenes with simple geometry. If you dont use DoF anothr approach would be to double or quadrouple Gi rays and boost LSM aswell, while keeping msi (Not to be confused with MIS) lower. This will massivly boost Rays/Sample and while your passes take longer, your image should be clean when you reach 15 passes.
That being said, everything is scene based anyways...

[maru]

1. am i right in thinking that if there's noise in reflection and/or translucency render elements, one needs set GI vs AA value higher?
2. is it ok to set LSM to lower than 1 (e.g. 0,5 - 0,2) if direct element cleans very fast?

1. I would say so, not sure about translucency, definitely about glossy reflections, not sure if this will work in all cases - sometimes reflections/translucency may probably need better AA sampling so lowering gi vs aa could help (I can imagine a situation like this).
2. I would say it's perfectly ok. If direct light renders nice, then why waste computing power? It will simply render faster with lower LSM.