Maxwell Render | Official public forum

so I was thinking... after voxelization of a scene: the scene has 2 separate rooms, voxelization for the first one (when it was ready to get rendered) took about 4 Gb of RAM. Then I modelled the second room and voxelization is now 8 Gb of RAM. But the rooms are totally separate from each other, with opaque walls. So can Maxwell understand that when preparing a scene - some of it can be omitted ?? instead of trusting the user to do some raw separation and then guess some more, can the engine itself decide - after short computation - what is need for the render and what has 0 or even close-to-0 impact on the results?

That kind of preprocess it's not that easy. Light rays are infinitesimal, and assuring that there's no little hole anywhere where a light ray can pass it's as heavy as the render itself.

But the way to go in this case is out-of-core. Having a 2 layers representation of the scene. At top layer you would have the objects, and low level where you have the geometry. And you only load low level under demand, so if no light ray hits the second room during render, the geometry of the second room it's not loaded.

This geometry out of core is a little more far away, but texture out of core should be something we can do in a reazonable time.

Before improving anything about that, I would like to see some improvement in performance while opening / importing high poly scene/models.

Tried some Evermotion Archinteriors scene and the importing process is slow as hell.

When isn’t crashing obviously

Anyway, would be nice one day having the possibility to exclude from rendering computation items not present in the camera view.

Probably doable for the texture, but I doubt for lighting rays bounces calculation.

Matteo Villa wrote: ↑

Mon Mar 15, 2021 8:43 pm

Anyway, would be nice one day having the possibility to exclude from rendering computation items not present in the camera view.

Well, any object potentially affects light of the scene even if it's not seen from camera. So again, the solution here is load geometry under demand.

luis.hijarrubia wrote: ↑

Tue Mar 16, 2021 10:16 am

Matteo Villa wrote: ↑

Mon Mar 15, 2021 8:43 pm

Anyway, would be nice one day having the possibility to exclude from rendering computation items not present in the camera view.

Well, any object potentially affects light of the scene even if it's not seen from camera. So again, the solution here is load geometry under demand.

What if a type of Clipcube (3D box or fence-line 'cube') could be defined around the imported model with only the entities inside the clipcube being used for light calculations and everything else outside it ignored? Our DataCAD has this type of feature to include/exclude geometry from views and makes it useful (and quick) to include/exclude what you need.

I know some people have implemented something similar in maxwell using python scripting. So they preprocess the scene and hide objects out of frustrum. But that changes the scene illumination in most scenes. Something like that can be done when setting the scene, and the user can take the risk os changing the visual result. But from inside the engine we can not do that, would be really biased. The unbiased way to solve it from engine should be geometry on demand.

let me give you an example to what I actually meant: take a typical apartment. your client wants to see the master's bedroom, the living room and the kitchen. on a typical apartment there's no way the kitchen cabinets affect lighting inside the bedroom. just to clarify my intension.

For something like that I usually create the entire project in Archicad.

Export room by room and then start the texturing one by one in Maxwell Studio if you need to lower the rendering time on older hardware or you want to have a slightly light project.

Then if the client request rendering where you can see troughs different rooms, or the project don’t have so distinct spaces, you will have to work on a new file with ALL the rooms present.


Developer could solve partially this problem providing a DIRECT LINK like in LUMION, so any changes in the Cad / Bim software will be seen in real time in Maxwell Studio.

But then they have to solve another problem.

Let’s say you’ve a WALL and you texture it in Maxwell STUDIO.

Then you delete the wall in Archicad.

You will see the wall deleted in Maxwell Studio too.
Great

But if you reverse the delete you will need to
re-texture it in Studio. Or they need to find a way to keep/preserve infos of the deleted items.

But that’s a minor problem, having a direct LINK between Maxwell Studio and Cad/Bim software would be great.


Another solution would be for developers, improve all the transformation tools present in Maxwell Studio, they need an upgrade for sure to give us more flexibility to model/transform/change the model directly inside Studio.

I'm not a fan of the linking method as I use Maya and there's no need for that. I just use display layers to turn off and on whatever is needed. then again, the whole idea I'm thinking about is to have things done automatically instead of the user having to figure out and test and re-test ...