OctaneRender™ Standalone v1.0 beta3.02 [obsolete]

[bepeg4d]

i have made a simple speed comparison between 3.01 and 3,02 on both, mac and pc and i must say that i'm a bit lost
on win 7 64 with 285 drivers i have a lot of problems with texture, as marcus said, probably i have to redo all the tests with newer drivers, infact the results are a bit strange, from +13,02% to -26,04%
on mac there is no issue with textures, the results are more stable but extremly worst, from -34,7% to -52,63%
in both the systems there is no time differences in the voxelization process
ciao beppe

[abstrax]

Ok, after the first feedback I think it's necessary to split the BRDF changes from the CUDA framework changes to see, where the main slow down happens. -> We should release some test builds (64bit only) with the old BRDF and different CUDA version combinations. This would also allow us to fine-tune the build settings, since we have now a higher flexibility in that area.

We will probably not get this done today, but hopefully tomorrow.

Please send me a PM, if anyone wants to contribute a test scene. It should be of moderate complexity and the copyright should allow us to put it on our web server for everyone to download. I guess we would need an interior scene, an exterior scene and maybe a studio scene. If there are no contributions, we will fallback to the demo scenes.

Thanks for your patience,
Marcus

[face]

Have seen the same effect in the latest build, but haven´t test it with this one.
On a scene with a render resolution of 1024x1024 pixels, the GPU memory uses ~30MB for the mesh and ~350MB for windows/desktop.
When the GPU is render, the GPU memory is changing between 380MB and 870MB.
I can also maximum render on a GPU with 2GB a scene which should fit in a 1GB GPU.
On a 1GB GPU a scene with ~100.000 tris is then the limit

face

[vimeo]http://www.vimeo.com/50630584[/vimeo]

[abstrax]

Have seen the same effect in the latest build, but haven´t test it with this one.
On a scene with a render resolution of 1024x1024 pixels, the GPU memory uses ~30MB for the mesh and ~350MB for windows/desktop.
When the GPU is render, the GPU memory is changing between 380MB and 870MB.
I can also maximum render on a GPU with 2GB a scene which should fit in a 1GB GPU.
On a 1GB GPU a scene with ~100.000 tris is then the limit

face

[vimeo]http://www.vimeo.com/50630584[/vimeo]

Do you have Windows Aero enabled? If yes, what happens, if you disable it?

Cheers,
Marcus

[face]

Do you have Windows Aero enabled? If yes, what happens, if you disable it?

Cheers,
Marcus

When i disable Aero, it jumped between 151MB and 631MB.

face

[abstrax]

When i disable Aero, it jumped between 151MB and 631MB.

face

Ok. Do you have any other 3D applications running on that GPU? What is kind of strange is that the memory report in Octane is stable and does not jump. Did you try to use the 1.6GB that are displayed as available VRAM in Octane?

Cheers,
Marcus

[face]

Ok. Do you have any other 3D applications running on that GPU? What is kind of strange is that the memory report in Octane is stable and does not jump. Did you try to use the 1.6GB that are displayed as available VRAM in Octane?

Cheers,
Marcus

Only Octane is running.
Verion 0.8 is stable, Verion 1.024 Beta 2.44 jumped around 120MB, Verion 1.024 Beta 2.57 around 230MB and Verion 1.030 Beta 3.02 around 480MB.

Haven´t test to fit the memory.

face

EDIT:
Maybe this are image buffers?

EDIT2:
The speed compared between the first and the last version is 25Ms/sec to 5Ms/sec
I know, the values are changed between the versions and can´t compare. But i got 45fps with Alpha 0.8