Linking and layers for render time

Can anyone tell me how much linking multiple project elements and/or having multiple layers affects render times? Basically, my scene starts outside a house; it’s fairly complex with lots of textures, trees, grass, etc. As the camera pulls back, it’s revealed to be a model/display inside a museum.

The “house display” is the most complex element of the scene as a whole. I’ve done as much instancing, material/texture linking, and decimating as I can. The only thing that allows me to even render without a crash is separating the grass/trees/road elements and the house itself into two separate projects, and linking them in another project.

What I’m wondering is whether, say, separating the scene into 5-10 separate projects and linking them together will help or hurt the render? Is there a point at which linking actually hurts the render? And additionally, would placing those linked sections into separate layers help? Is appending a better alternative?

Aside from grouping (which I’m doing to each project element), if I’m missing any render saving methods please let me know. I’m reluctant to do any baking, due to the complexity of the scene and how it looks with camera movement. Insight is much appreciated!

So far as I have ever yet experienced, “library linking is painless.”

However, as I read your description of “what you are ultimately trying to achieve,” I would kindly suggest that you should probably approach this sequence as “two-or-more separate and distinct shots,” joined (in the editing step) “by artful ‘cuts.’”

Furthermore, I suggest that you should decide … and commit … to “exactly where that ‘cut’ is going to happen.” Do this by:

  • Create OpenGL Preview renders, then …
  • Edit these together, using VSE or the external video-editor of your choice, then …
  • Armed with now-exact(!) information of just what frame-numbers you will actually need, render those frames.

If you observantly watch “actual films or TV episodes,” you will always find that they used “several ‘cuts.’” And, you’ll realize that they did this for two distinct reasons:

  • “Practical necessity.”
  • To save time, and keep the story moving.

The first of these considerations are based on technical realities such as the one that you are now facing. But, the second is an editorial, and cinematographic, therefore “artistic,” choice.

In your OpenGL Preview “animatics,” you should independently produce several “long strips of film.” Then, you should put on your Editor’s hat, and “cut the hell(!) out of it,” until what you come up with is as tight(!) as it can possibly be.

(This is the point at which you should aggressively exercise your “editorial and cinematographer” sensibilities. “OpenGL Preview” ‘footage’ is basically free.)

Only after convincing yourself that “it is truly not possible to ‘cut’ this thing still-further” should you commit to more-serious render time.

In conventional film-making, “footage is free.” It takes exactly one calendar minute to expose one minutes’ worth of film. Therefore (think ‘Waterworld’ …), some directors exposed miles of the stuff, then struggled to cut it into a presentable picture “in the editing room.”

CG, however, is precisely the opposite: “every miserable frame comes at a price.”

Therefore, whereas the conventional dogma is “shoot, then edit,” the (present) CG reality is: “edit, then shoot!”