Loading...
> >

Splitting the render in two times

LenioTGLenioTG Posts: 2,118
edited December 2018 in New Users

Hi everyone :D
I've noticed that, even using always my CPU to render, the total render time is much smaller if I divide the scene in 2 parts: one with the characters, another with the other stuff.

  • If I render everything together, in 40 minutes I get 200 iterations (5 iterations/min).
  • If I render the characters alone, in 10 minutes I get 500 iterations (50 iterations/min).
  • If I render the other stuff alone, in 10 minutes I get 100 iterations (10 iterations/min).

So, the complessive time for the other stuff is the same (2x iterations/min for half the stuff), even if you can blur it more, and so you need less iterations.

But the difference on the characters part is huge! As I've said, I'm using just my CPU.

Why is that so? Am I doing something wrong?

PS: I'll attach the renders I'm talking about :)

127 char.png
2560 x 1440 - 2M
123.png
2560 x 1440 - 7M
Post edited by LenioTG on

Comments

  • Richard HaseltineRichard Haseltine Posts: 100,842

    If there's no scene then almost any path that bounces off the characters is going to go to infinity and terminate, so it's much quicker than with the environment there. Taking the characters out of the environment alows some paths that would have hit the characters to escape to infinity (assuming it isn't enclosed) and in any event almost certainly simplifies the interactions (the enviroment probably isn't using things like SSS) so even those paths that do bounce around repeatedly are quicker to calculate.

  • LenioTGLenioTG Posts: 2,118
    Richard Haseltine said:

    If there's no scene then almost any path that bounces off the characters is going to go to infinity and terminate, so it's much quicker than with the environment there. Taking the characters out of the environment alows some paths that would have hit the characters to escape to infinity (assuming it isn't enclosed) and in any event almost certainly simplifies the interactions (the enviroment probably isn't using things like SSS) so even those paths that do bounce around repeatedly are quicker to calculate.

    Oh, that's why!

    Now I see, thank you! :D

  • DekeDeke Posts: 1,631

    Works great for animation. If the camera is locked down, then you only need one frame of the background. Works if there is no interactivity needed, for example shadow of moving figures falling over the background items. With animation you render as a PNG sequence so that you have an alpha layer and then matte the layers together in a program like Adobe After Effects.

  • LenioTGLenioTG Posts: 2,118
    Deke said:

    Works great for animation. If the camera is locked down, then you only need one frame of the background. Works if there is no interactivity needed, for example shadow of moving figures falling over the background items. With animation you render as a PNG sequence so that you have an alpha layer and then matte the layers together in a program like Adobe After Effects.

    I still haven't learnt animations!
    I know how to use Premiere, but not After Effects!

    Thanks for the tip :D

Sign In or Register to comment.