Can a SINGLE object be split into multiple layers to save GPU memory

I’ve got a building that consists of a main object …(the building) than lots of other objects … that go with the scene. Currently I have all those “other” objects on different layers so that if these other options are not in the cameras view, I just uncheck those layers. This is all due to wanting to render with GPU ( I only have 1 gb memory).

However, even doing this, the main building is still too heavy polycount to render with GPU.

I am currently using a mirror modifier on the main building if that makes a difference. I’m not going to be able to apply it because I still have additional modeling to do.

But my question is this… Can I somehow split this main object up and split it up using 4 layers…so select a quarter of the verts… move to a layer … select another 1/4 of the verts and move to it’s own layer ect… So if my camera is looking only at 1 portion of the building… I can turn off those other layers and then be able to render with GPU

This is where instancing is useful rather then the mirror / array modifiers… pretty much those modifiers generate extra polygons… whereas instances just reference the first and dont take up any space (well they do, like 4kb or something small like that)

Pretty much, click on your object, and press alt+d, that will create a linked duplicate (aka instance) do you have any other objects which are the same in your scene? if you do make sure to take advantage of instancing rather then just normal duplicating.

rendering it in 4 lots means you wont get any indirect lighting between the four renders… it is possible to do it but its alot more work in the comp

How much of the back side will be out of view completely? You can experiment with using mask modifiers to vert groups, and turn on and off mask modifiers for rendering only parts of an object.

Hey Craig and doublebiship… thanks for your input… maybe if I showed you what I am doing things might be easier to visualize. Take a look at this link to my project.

If you scroll down, you will see some different shots of the Harley Hotel & Casino :slight_smile: , there is also some shots on the previous page

The rooftop pools/water and pool area entranceways and fences are on seperate layers. The columns, road, and some minor stuff are on seperate layers… so if I’m shooting a render with a camera angle that doesnt show the pools, I just dont have that layer on.

@doublebishop … I’m actually using instanceing now on some objects (not the main building) and I have definitely learned the value of these. But the main hotel building is all set up using mirrors so I really can’t change that right now.

@Craig I will have to investigate the mask modifiers your refering to. I have no idea what those are or how to use them.

I did a little experiment… took a cube, subdivided bunch of times. I was hoping I could just select some verts … like the back half of cube… and move to a new layer. But that doesnt work. I would have to seperate the cube into different objects. and then move each part of the cube to a different layer.

Basically on those scenes that I am going to be showing the hotel in a wide shot… I’ll just have to bite the bullet and render with CPU. But many shots will be closeup on different parts of the hotel, So much of it won’t be seen, on those shots Im trying to figure a way to break up that hotel into about 4 or more render layers.

Select a quarter of your model you want to hide, add that to a new vertex group and name it. Add a Mask Modifier and name that vertex group, and turn it on in edit mode and you will see that portion disappear - if not, you may need to check invert. Do this for each section of the model, and then you will have the ability to turn off and on the visibility of your model without trying to split into objects and layers.

Thanks for running me through it Craig… With the understanding that the main reason for doing this is to reduce GPU memory usage for any given scene… Do you know if cycles will actually use less GPU memory when rendering the scene when you turn off the visibility? I guess my other question would be… would I need to apply the mirror modifier, before doing the mask steps you mentioned?

If you have a mirror modifier on, then the groups masked will reflect across the center axis. The hidden geometry does make rendering faster, I already use this for parts under my models that are out of camera. The only problem here is that if your object has a subdivision modifier on it for smoothing, the edges that are loose might be affected by the lack of geometry.