OctaneRender™ 4 RC 5

[funk]

Sheen is a lot darker on the universal material compared to metallic and glossy materials

universal_dark_sheen.png

Hi,

Universal material is actually correct here as it has been renormalized to avoid energy gain, this will be fixed on both glossy and metallic material to match universal material in next version.

Cheers,
Wallace

OK thanks Wallace. I think it's important we get a beter match across all materials too.

I do feel the sheen on the universal material is too hard to see in most situations though. In some cases I'm even adding a multiply node after my color node to boost the color.

[abstrax]

...
Ok so where does this leave us? Paul told me he didnt know how to proceed and asked me to go to OTOY.

Could you send the Modo scene to me? I will try to reproduce the issue. I will add logging of a hash for the relevant attributes to see which attributes change between frames. I hope I can then give a more definitive answer to where the problem is.

I actually attached a zip of the scene (plus 2 orbx exports) to the post you replied to earlier viewtopic.php?p=347354#p347354 but I'm attaching a modified scene here too

At the moment it looks like the issue is in the Modo API.

I agree it could be a problem on the modo side, and I assumed Paul would be able to output some kind of debug info pinpointing where the problem was. I actually asked him about this and he said he didn't know how to test it.

Before you send the scene, could you remove all objects except the vertex motion blur one and try again? If the issue is still there, please send that reduced scene to me. Thanks a lot in advance.

A modified scene is attached. This one seems to reproduce the "random" problem a bit easier, since the deformation is running over more frames.

I have included 2 orbx exports from this scene. One has working motion blur, the other doesnt. The motion blur is seen best around frames 19-21.

I also did an ABC export from modo's own exporter just in case it helps. As I said earlier, this alembic always works, but ngons don't get triangulated, so you can end up with bad rendering in octane (although its not obvious in this example).

I tried it in Modo 11 and added a bunch of logging to see what is being exported and the polygon-to-vertex indices are pretty much alternating between two versions for every exported frame. So it's most likely an issue with the data provided by Modo to the plugin. I.e. there isn't much we can do about it unfortunately. I will ask Paul to follow it up with Luxology.

[funk]

I tried it in Modo 11 and added a bunch of logging to see what is being exported and the polygon-to-vertex indices are pretty much alternating between two versions for every exported frame. So it's most likely an issue with the data provided by Modo to the plugin. I.e. there isn't much we can do about it unfortunately. I will ask Paul to follow it up with Luxology.

Thanks for looking into this Marcus.

I guess Paul just got very lucky that some exports had constant topology. It seems the longer your animation is (more frames), the greater chance there is it wont work.

This presents a big problem for modo users. Exporting to orbx was the official recommended method for "working" vertex motion blur.

[mikinik]

in octane, everything is as usual, no changes... In all the latest builds alembiс not rendered.

don't worry, this is the last time i've run octane. bye.

[abstrax]

in octane, everything is as usual, no changes... In all the latest builds alembiс not rendered.
...
don't worry, this is the last time i've run octane. bye.

Well, we are not aware of any Alembic issues. I also haven't seen bug reports from you during the last couple of months either. Did you contact our support team?

Anyway, if you would like to get help: To investigate (and hopefully fix any issues) we would need the exported Alembic file, the method how it was exported and if possible the original 3D scene. Thanks.

[funk]

Hey guys, someone got in touch with me about denoiser crashing on his 8k image.

I had a look at this older XB thread, and it seems we should be able to denoise any res octane supports: viewtopic.php?f=33&t=67492

Added tiling and multi-GPU support for the AI Denoiser. You can now denoise any resolution up to Octane's maximum, and the Denoiser will consume less memory per device (approximately 450MB per device maximum).

I tried denoising in modo at 8K res. The scene just contains a single cube. The render takes ages to appear, and system RAM usage hits 12GB, which seems unusually high (RAM usage is about 4GB without denoiser). Why is system RAM usage so high?

I don't get a crash, but the denoiser fails with this in the error log:

Code: Select all

CUDA error 2 on device 0: out of memory
  -> failed to pin memory
denoiserThread0: Denoiser thread failed to pin a buffer. Restart required

Im using a 1080ti with 11GB. The denoiser works at 4k. Is this a bug?

EDIT 2: I also tried in standalone rendering a geometry > "plane" item at 6k with denoiser. The same problem occurs. GPU-Z does not show my VRAM anywhere near full, so this must be related to system RAM?

[mikinik]

abstrax , we discussed this problem three years ago - viewtopic.php?p=229816#p229816 later, the conversation about this was raised again by other users - viewtopic.php?f=33&t=65518
directly in 3ds max I can open alembic without any problem, not only in 3ds max, Maya, modo and everything there is OK except for the octane...
I have not used octane for a long time, I decided to check if something has changed, but no.

[abstrax]

Hey guys, someone got in touch with me about denoiser crashing on his 8k image.

I had a look at this older XB thread, and it seems we should be able to denoise any res octane supports: viewtopic.php?f=33&t=67492

Added tiling and multi-GPU support for the AI Denoiser. You can now denoise any resolution up to Octane's maximum, and the Denoiser will consume less memory per device (approximately 450MB per device maximum).

I tried denoising in modo at 8K res. The scene just contains a single cube. The render takes ages to appear, and system RAM usage hits 12GB, which seems unusually high (RAM usage is about 4GB without denoiser). Why is system RAM usage so high?

I don't get a crash, but the denoiser fails with this in the error log:

Code: Select all

CUDA error 2 on device 0: out of memory
  -> failed to pin memory
denoiserThread0: Denoiser thread failed to pin a buffer. Restart required

Im using a 1080ti with 11GB. The denoiser works at 4k. Is this a bug?

EDIT 2: I also tried in standalone rendering a geometry > "plane" item at 6k with denoiser. The same problem occurs. GPU-Z does not show my VRAM anywhere near full, so this must be related to system RAM?

denoiser_crash.png

For the the denoiser to work we are rendering quite a few more passes (which you don't see), which require a lot of system memory. The denoiser then needs to load them to the GPU. The way the denoiser works right now the channels of these render passes need to be pinned, which can fail if it's too much.

-> It's a limitation, but we are currently refactoring this area, which should remove the requirement to pin and release an update when that's done, but it will take a few of weeks to have this properly finished.

[abstrax]

abstrax , we discussed this problem three years ago - viewtopic.php?p=229816#p229816 later, the conversation about this was raised again by other users - viewtopic.php?f=33&t=65518
directly in 3ds max I can open alembic without any problem, not only in 3ds max, Maya, modo and everything there is OK except for the octane...
I have not used octane for a long time, I decided to check if something has changed, but no.

Thanks for getting back to me. I made a small change in the Alembic loader which hopefully solves this problem in the next version. You can find a long explanation here in the thread of Jolbertoquini.

[mikinik]

abstrax, thanks!