OctaneRender™ 2.1 for LightWave™ - Daily builds

[FrankPooleFloating]

Yes J, thanks so much for fixing the IPR zoom/scroll etc.

Also heads up Juanjo: although you write (and presumably speak) English as if it was your native tongue, you consistently say one thing that always sticks out, and blows your cover: "Thanks for report it".... Thanks for reporting it - is what you want.

[MrFurious]

nice catch Frank

[gordonrobb]

Thanks for report it Frank

[juanjgon]

Yes J, thanks so much for fixing the IPR zoom/scroll etc.

Also heads up Juanjo: although you write (and presumably speak) English as if it was your native tongue, you consistently say one thing that always sticks out, and blows your cover: "Thanks for report it".... Thanks for reporting it - is what you want.

Thanks for reporting it

My English isn't very good, I am glad that at least all people can understand me

-Juanjo

[MrFurious]

Yes J, thanks so much for fixing the IPR zoom/scroll etc.

Also heads up Juanjo: although you write (and presumably speak) English as if it was your native tongue, you consistently say one thing that always sticks out, and blows your cover: "Thanks for report it".... Thanks for reporting it - is what you want.

Thanks for reporting it

-Juanjo

... it just doesn't have the same Juanjo vibe

[Godstrike]

Please see attached. Looking at my nodes while making the test scene, I see the problem presents itself when you are mixing HDRIs.

Carlo, the problem is the Mix Texture node. The issue is that the render target nodes (both Texture and Daylight environment) only supports the texture rotation transformations if the Image Texture node is directly linked to their texture inputs. If you put any node between the image and the render target nodes, the rotation parameters don't work.

You only need to remove the texture mix node to be able to rotate the HDRI backgrounds.

-Juanjo

Thanks Juan, yeah I figured as much when I was making the test scene:)

Any chance that the heading info could be piped through the intermediate node in a future release?

Carlo

[juanjgon]

-this is old problem but it is so serious that I now make the max. samples to use way over needed and then always do Abort+Cont because if I let the render go to the end, LW gets in not responding state often (the 99% problem), not always but too often.
Antti

Antti, I am testing your 99% problem here with your render target file and I am not able to reproduce it ... This problem can be something wrong with your system or your scenes.

Today or tomorrow I am going to release a new daily build (2.13.0) with the last Octane API core. Please test your problem with this new build to see what happens. If this build doesn't fix your issue I am going to need one of your full scenes here to test it.

-Juanjo

[juanjgon]

The daily build has been updated. The current version is 2.13.0.

Remember to update your NVIDIA drivers. This release uses CUDA 6.5 and needs the last drivers to work.

-Juanjo

[MrFurious]

Muchos gracias. One question regarding this build being optimized for the newer Maxwell CPU's.. will this effect the performance of older GPU's, namely my gtx590/580?

[juanjgon]

Muchos gracias. One question regarding this build being optimized for the newer Maxwell CPU's.. will this effect the performance of older GPU's, namely my gtx590/580?

From the standalone release notes:

"Until now, Octane kernels were compiled with CUDA 4.0, which provides a compiler based on Open64. With CUDA 4.1 NVIDIA introduced a new compiler based on LLVM, which produced totally different output and made Octane 10-15% slower. On the other hand, GPUs of newer compute models (like Maxwell) couldn't be built with CUDA 4.0. So we had to use a mix of different toolkits to support the various GPU architectures. Another problem was that on Mac OS you could either use only CUDA 4.0 on Xcode 3 or CUDA 4.1 (or higher) on Xcode 4 (or higher), effectively prohibiting the toolkit mix we used on Windows and Linux (which is the reason why OSX builds didn't have Maxwell support yet).

To fix all that mess and to allow us to do some interesting things in the future, we really needed to switch to a recent toolkit. It took us 3 tries to adapt the code for CUDA 6.5 without a too big performance loss. Since the output of the LLVM based compiler is very different from the old Open64 based compiler (CUDA 4.0), most scenes will render with slightly different speeds compared to version 2.12.1. Some scenes will be faster while others will be slower. It's hard to say what exactly works better/worse, but as a general trend, scenes with instances (i.e. most scenes) often render a bit faster in DL and PT, while single mesh scenes (like the old benchmark) render a bit slower in DL and PT. PMC seems to be a bit slower in most cases. Maxwell 2 GPUs got a lot faster (because of the switch and other optimizations)."