Optimization Techniques

[snivlem]

Hi, sometimes I'm requested to some pretty heavy scenes with walkthroughs, unfortunately sometimes I cannot "cheat" by cutting down down on the polycount and have to show the whole model (lots of glass and overhead views with lots of detailed landscaping).

Now I have some questions regarding V-Ray.

1) Is it better to have say, a building, is it better to have this as one large Polygonal object (everything combined). Or is it better to have the building separated into tiny pieces (ie window is a separate object to a wall, each wall individual etc). I guess the main question is it better to have one large object or many small objects (thousands)?

2) When animating in this large scene, if I have a camera that's only showing a fraction of the scene, is it better to delete the rest of the scene? Or will V-Ray automatically discard anything that is not in the clipping plane and reflection/refraction rays?

3) Will extremely large faces (ie physical area) slow V-Ray down? compared to the same surface area with many smaller faces?

4) Do large bounding box scenes slow V-Ray down? (when it gives this warning: V-Ray warning: Scene bounding box is too large, possible raycast errors.)

5) With tiled exrs, is it best to have them in factors of 2? 256x256, 1024x1024 etc?

[Zdravko]

1) Is it better to have say, a building, is it better to have this as one large Polygonal object (everything combined). Or is it better to have the building separated into tiny pieces (ie window is a separate object to a wall, each wall individual etc). I guess the main question is it better to have one large object or many small objects (thousands)?

You can use whichever approach you like, they will not affect the render time significantelly.
3d artists usually combine objects with similar materials: glass, window frames, concrete, etc. for better control.

2) When animating in this large scene, if I have a camera that's only showing a fraction of the scene, is it better to delete the rest of the scene? Or will V-Ray automatically discard anything that is not in the clipping plane and reflection/refraction rays?

V-Ray renders only what the camera sees.
If you want to have correct reflections it's better to leave the whole scene. On other hand the export may be slower and it may take more memory.

3) Will extremely large faces (ie physical area) slow V-Ray down? compared to the same surface area with many smaller faces?

Large faces will need less memory , so rendering should be faster.

4) Do large bounding box scenes slow V-Ray down? (when it gives this warning: V-Ray warning: Scene bounding box is too large, possible raycast errors.)

Try to keep them whit a reasonable size. To large bounding box may cause numerical errors. Such errors may slow down V-Ray a bit.

5) With tiled exrs, is it best to have them in factors of 2? 256x256, 1024x1024 etc?

No, there will be no significant difference.

[snivlem]

Thank you so much. They're probably basic questions so I appreciate the time you took to reply to that.

I've always wondered about these!

[vlado]

In addition to what Zdravko wrote, I would like to make a note on the second question.

2) When animating in this large scene, if I have a camera that's only showing a fraction of the scene, is it better to delete the rest of the scene?

If the out-of-view geometry does not contribute significantly to the rendering, it will be better to hide it, yes.

Or will V-Ray automatically discard anything that is not in the clipping plane and reflection/refraction rays?

No, V-Ray does not discard anything (when you have GI, reflections and refractions, it is very hard to predict which parts of the scene will be needed). However, V-Ray will attempt to delay the calculation of some things as much as possible (e.g. loading of proxy objects, displacement, VRayFur etc).

Best regards,
Vlado