OctaneRender™ Standalone 2.21.1

[riggles]

I've been on a ~6 month sabbatical from 3d, and I have to say, the amount of work that's been done on Octane is impressive. It's good to take a step back and get some perspective, because up close it's easy nit-pick and get impatient about when certain feature requests are going to be implemented and indignant about why certain functions are missing. But really, look at how much has been accomplished in the last 6 months, and then look back to where things were a year ago. Real progress has been made.

Cheers~!

[manalokos]

When changing the focal lenght of the camera the aperture goes from 0 to some other value, I almost never use the aperture, so it is quite bothersome to have the camera turning on depth of field just because I change the view angle.

The other options of sensor size and focal lenght are great anyway.

[whersmy]

no way!

octane is even faster with 2.21?? I worked for months on 2.12 due to the new cuda toolkit 6.5 and lower performance on older cards due to maximum compatibilities with newer cards
but i`m checking now:
benchscore PT on 2.12 -> 24ms
benchscore PT on 2.21 -> 30ms

that`s some crazy coding

[RobSteady]

octane is even faster with 2.21??

You have set coherent ratio to 1.0 in 2.21, that's why.
Coherent checkbox switched on in earlier versions equals a value of 0.35 if I'm not mistaking...

[smicha]

octane is even faster with 2.21??

You have set coherent ratio to 1.0 in 2.21, that's why.
Coherent checkbox switched on in earlier versions equals a value of 0.35 if I'm not mistaking...

That is right. Anyway 580/590 are still powerful, yet power hungry cards.

[rappet]

octane is even faster with 2.21??

You have set coherent ratio to 1.0 in 2.21, that's why.
Coherent checkbox switched on in earlier versions equals a value of 0.35 if I'm not mistaking...

That is right. Anyway 580/590 are still powerful, yet power hungry cards.

Yes, 'old' 580 card is pretty powerful, bu not compared to 780 and Titans.
The 780Ti gives about double renderpower copmared to 580 classified.

I swapped my powerful quad 580 with triple 780 and compared benchmark results (Octane 1.20) in the same case, so therefor same airflow. Big difference is that triple 780Ti have a slot space between the cards and the 3x 580 classified were sticked together.
3x 780Ti calculated down per 1GPU: 8.10 Ms/sec
4x 580classified calculated down per 1GPU:: 4.18 Ms/sec

greetz,

[smicha]

That's right. 780 (not Ti) is about 70% faster than 580, much cooler, and with 6GB is just great.

[roeland]

The RAM usage limit (by default set to 4GB) tells Octane how much RAM it can use for out-of-core textures. The bar graph in the preferences shows approximately how much memory is being used on your system (as reported by the operating system), you can use that as a rough guide as of how much you can increase the limit.

The out-of-core texture memory is page-locked and thus unavailable for other applications. This means eg. that if you have 16GB of RAM, and Octane uses 6GB of out-of-core textures, other applications will behave as if your computer has only 10GB of RAM. Anyway since the GPU can access the video RAM much faster than the system RAM, I would expect a lot of speed loss if Octane needs to use that much out-of-core memory.

The GPU headroom is there because CUDA needs to have some free memory available to launch the render kernels. The default value, 300MB should be enough to start rendering. You don't need to change this value unless you frequently see kernel failures when Octane starts using out-of-core memory.

When octane starts using out-of-core memory, you can see this in the memory statistics. Octane will show a second bar graph (coloured yellow) showing how much out-of-core memory is being used.

--
Roeland

[coilbook]

Will we see better static noise in future releases? Like glossy reflection lights that are static etc
100% static would be nice like it was rendered using CPU
Thanks

[abstrax]

Will we see better static noise in future releases? Like glossy reflection lights that are static etc
100% static would be nice like it was rendered using CPU
Thanks

It's on the to-do list, but I can't say yet when it will be implemented.