Reality OpenCL vs CPU acceleration

[White-B]

I have tested yesterday the new driver for Nvidia: 411.63
It kind of works with Reality, Luxrender for OpenCL rendering.

I use the texture collection feature with 2000 pixels.
For the scene confugration i let it as is for opencl rendering:
Open CL group size: 64
Type: Mono-Directional
FFs Removel: 1
Sampler: Sobol
Max Path leght: 16
Light strategy: simple

Here is a picture without opencl

her is one with opencl

i haven't tried much, but it renders at least 4 times faster compared to cpu only renders.

my specs:

AMD FX 8350
NVIDIA GTX 1060 6 gb MS Armor edition
16 gb of ram

The resolution is 20560 x 1440.

[sigstan]

Well, maybe NVIDIA is finally getting their act together on the driver side

I hope you mean a resolution of 2056 x 1440 and not 20560 x 1440. Otherwise I'm even more impressed

In the comparison on the CPU-version I was wondering, if you had the same setting for type/sampler for the CPU-version (mono-directional/Sobol)?

How long did it take in the OpenCL version, before Luxrender actually started rendering something to the screen (kernel compiling)?

PS:
You can set FFs Remove to zero, since it is probably not going to have any other effect than taking up more memory.
Also Light Strategy is better set to Auto than Single, unless you only have a single light source in the scene.

[Marcel2586]

[White-B]

I mean 2560 x 1440.
After some tests i can say that Sobol Bi- and Monodirectional Sobol work. It usually takes 1 to 3 minutes.
Thanks for the suggestions. I had the basic settings for this test render and single was the standart setting. Atm i am rendering with the Bi-Directional setting.