MR proxy or simple maya instance?

[eem]

From my testing mip_binary_proxy runs exactly as instances do, only they are read from disk and can be unloaded.

If your in 2009, mip_binaryprox and BSP2 don't work together. But this appears to be fixed in 2010.

Assembly proxies (The autodesk implemented render proxy function) in 2009 was rubbish, used considerably more ram and took longer to render than instances. I have not tested them in 2010 though, so they might be different.

[royterr]

From my testing mip_binary_proxy runs exactly as instances do, only they are read from disk and can be unloaded.
.

so just like .maps textures, they dont use RAM?

[eem]

so just like .maps textures, they dont use RAM?


I don't know. I think it has to load and unload the entire object. But I'm not sure.

But the memory the model takes up in memory is actually the least of your worries. I have about eighty 500,000 poly trees in a scene right now , the tree itself is maybe 60 mb of mesh data, but it's using up 3.3 gb for the BSP structure with BSP2! That means each tree is taking up 40 mb for the BSP structure.

Ya thats better than 40 mb + 60 mb for the mesh. But still. That is unaccpetable. Honestly, I don't think you could call these real render proxies until a proper BSP acceleration method is integrated, as the memory savings is absolutely nothing compared to other renderers.

And large BSP seems to not even function in my current scene. It craps out about half way through.

[royterr]

I don't know. I think it has to load and unload the entire object. But I'm not sure.

But the memory the model takes up in memory is actually the least of your worries. I have about eighty 500,000 poly trees in a scene right now , the tree itself is maybe 60 mb of mesh data, but it's using up 3.3 gb for the BSP structure with BSP2! That means each tree is taking up 40 mb for the BSP structure.

Ya thats better than 40 mb + 60 mb for the mesh. But still. That is unaccpetable. Honestly, I don't think you could call these real render proxies until a proper BSP acceleration method is integrated, as the memory savings is absolutely nothing compared to other renderers.

And large BSP seems to not even function in my current scene. It craps out about half way through.

I totally agree and this is what InTerceptoV (http://forums.cgsociety.org/member.php?u=93456) from blur studios pointed out.
what's the use of MR proxys if they are going to be loaded in ram anyways to do the ractrayce/GI/Bsp calculations.

I am beginning to really wonder what's the the use of MR proxys.

[Hamburger]

Are you guys rendering with final gather calculating?

On my 9GB machine the renders would always die on me even on BSP2 due to lack of memory.

However there's one method you could try....if you calculate FG without trees, then unhide the trees after FG is calculated....then freeze your FG and render, Your computer should be able to do it without too much hassle.

I'm not certain this will work, but it did work for me during one project. Background trees don't really need FG anyway.

[tostao_wayne]

From my testing mip_binary_proxy runs exactly as instances do, only they are read from disk and can be unloaded.

If your in 2009, mip_binaryprox and BSP2 don't work together. But this appears to be fixed in 2010.

Assembly proxies (The autodesk implemented render proxy function) in 2009 was rubbish, used considerably more ram and took longer to render than instances. I have not tested them in 2010 though, so they might be different.


have you tryed mip_binaryprox in 2010? how can i activate it? i would like to do some test with it and maya 2010.

[eem]

have you tryed mip_binaryprox in 2010? how can i activate it? i would like to do some test with it and maya 2010.

Same way you activate them in the previous version of maya

http://mymentalray.com/wiki/index.php/Production_shader_library


and royterr, as you prolly saw tostao pointed out in my forest thread, all my issues with the mib_binaryproxy were coming from me using scanline rendering. When you turn scanline off you can do thousands of proxies easily. The same is true for instances.

I just tried with the MR Proxy Assemblies in 2010 and it doesn't work. 2000 trees with mip_binaryproxy takes 700 mb. 80 trees with the assembly proxies takes 6 gb! Autodesk really screwed up on there assembly integration. Unless I'm also missing something there.


On a sidenote, does anyone know how to get per object antialising seetings to work on the mib_binaryproxy?

[JayHoo]

On a sidenote, does anyone know how to get per object antialising seetings to work on the mib_binaryproxy?

The problem here is that the mip_binary_proxy uses the transform node settings (mental ray / flags). The assembly proxy uses mesh settings (render stats etc). :banghead:

[eem]

The problem here is that the mip_binary_proxy uses the transform node settings (mental ray / flags). The assembly proxy uses mesh settings (render stats etc). :banghead:

Yeah that is the problem.

I was thinking, maybe you can stick the per object sample settings on the transform node in some way.

Any node can pass extra attributes to mental ray I'm pretty sure.

would anyone by any chance know the mental ray commands for per object AA settings?

[eem]

The code put in the object section of the .mi file is

samples 0 2

Does anyone know how to put that in the transform node so it exports with mental ray?

[InTerceptoV]

Hey guys... when you are doing a lot of instancing, make sure you use BSP2, and turn off any kind of scanline. This will save you a lot of memory. Rely only on raytracing.

Proxy's can be flushed out of memory when not needed, so my main point was if you have a lot of geometry that's heavy and noninstanced you can use proxy, but if you have just a few trees that you will instance 5000 times or more, you probably don't need proxy feature.

BSP2 is the big factor in everything if you are instancing, not really the proxy.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.