On 03/15/2018 05:53 PM, Norbert Kern wrote:
> Alain <kua### [at] videotronca> wrote:
> 
> You are right with the intervals 1 issue.
> If it wasn't a Gilles Tran code, i'ld have checked more carefully...
> 
> Meanwhile I rendered with intervals 1 and samples 70.
> It's not faster, but the result is slightly better.

Thanks for running again this way and reporting the speed! Your result 
lines up with my recent experience and what I see in the media code as 
I've been trying to digest it. In the code, intervals and samples end up 
both combining to drive the number of initial samples taken.

If you need 70 initial samples to resolve the smaller/rapid changes in 
your cloud or your shadows - you need them - no matter the intervals / 
samples combination to get them and the performance won't be much 
different due the initial specification. In some ways needing a higher 
number of initial samples with a media is like needing a high 
max_gradient value with isosurfaces due the function used.

In any case, intervals 1 is the way to go with method 3 for quality 
always - and often enough the 'density/shadow function' will be such 
that the adaptive method 3 with intervals 1 and a low samples setting is 
much, much faster.

Bill P.

Post a reply to this message