OTOY Forums

And is 16000 samples and 16 bounces (for this scene) not overkill?

octane rc2 integration 1h6min (448 fermi cores) pmc5k maxdepth16 filter 1.2

vray 1h23min ([email protected]) bruteforce64+lc3000s screen 0.01px+prefilter+filternearest10+caustic 6000sampes (36000000) dmc aa1-4 noisetrash 0.005

does Betrand know you've nicked his lunch?

enricocerica wrote:

gabrielefx wrote:vray scene is rendered with cpu

all materials have normal maps

all materials are converted one by one respecting all parameters

vray render took 15 minutes with one single 3.2Ghz processor

Octane for Max rendered for about 50 minutes, PMC 16 bounces 16000 samples, 4 GTX580

What really surprises me is the rendering time, are you sure about 15 min on a single 3.2Gh processor with micro-displacement vs 50 min on 4 GTX580 in non indoor environment

yep!, seem quite strange such time (gtx580 x4), Gabriele, did you try to see how was coming out the render after few minutes?, why the choice of 16 bounces?

The artist who did the scene used vray optimized parameters...with not so high values.

To compare vray with Octane rendering time I should use bruteforce for primary and secondary rays.

It's a really simple scene with detailed objects, one box with two walls opened, 3 planar lights and 4 flat deflectors...simple white boxes near the objects.

No matter what's the rendering time but the automatic translation of vray materials that is wrong.
Max color correction is not translated
Vray two sided material is not translated
black and white bumps are treated as color image textures and not as float
Vray normal maps aren't translated or go in the wrong slots
Vray camera is not translated then you have to create your camera and adjust the exposure

I sent the scene to Karba

Unfortunately on the web we don't have 3d libraries optimized for Octane for Max, then we have to conver them manually., and the process is not easy...but fun (for me).

gabrielefx wrote:The artist who did the scene used vray optimized parameters...with not so high values.

To compare vray with Octane rendering time I should use bruteforce for primary and secondary rays.

It's a really simple scene with detailed objects, one box with two walls opened, 3 planar lights and 4 flat deflectors...simple white boxes near the objects.

No matter what's the rendering time but the automatic translation of vray materials that is wrong.
Max color correction is not translated
Vray two sided material is not translated
black and white bumps are treated as color image textures and not as float
Vray normal maps aren't translated or go in the wrong slots
Vray camera is not translated then you have to create your camera and adjust the exposure

I sent the scene to Karba

Unfortunately on the web we don't have 3d libraries optimized for Octane for Max, then we have to conver them manually., and the process is not easy...but fun (for me).

You're right Gabriele, unfortunately too many V-ray materials are not translated (or wrong translated), we all hope that soon this problem will be solved

15 min in Vray You get with BF (brute force on both primary & secondary) = unbiased?
Or all this was left with Bertrand Benoits settings? (If so, that's bit unfair comparison.. =)

I don't think this was meant as a speed comparison. I doubt you'd need 16 bounces and 16,000 samples to get a noise-free image.

Regarding image quality, I prefer the VRay image because it's sharper and appears to have more detail. I don't think we can control the AA filter Octane uses, so that's probably where the difference stems from. Or maybe unbiased renderers don't have AA filters and due to their nature require some sharpening in post?

Chris_TC wrote:I don't think this was meant as a speed comparison.

Probably You're right..

I love comparisons, but..Here's the main problem I see here:

You take two engines, Gab: one that has 10+ of development and the other one that isn't even finished yet.. Then You compare quite hiEnd model that was made in mind of program's (vray engine) possibilities..with things like displacement etc..

It's somehow akward to compare these in this way =) - if You get few millions polygons sculpt without any displacement, dress that with HDR..put under same conditions (both engines workin' in unbiased mode) and then compare..in that case it would be a bit more interesting.

Now You comparing stuff when You know that it's unfair to compare at all.. No offense, but this feels strange that's all =)

glimpse wrote:15 min in Vray You get with BF (brute force on both primary & secondary) = unbiased?
Or all this was left with Bertrand Benoits settings? (If so, that's bit unfair comparison.. =)

no, lightcache screen 0.00px Progressive Path Trace on both primary & secondary = 95% unbiased, 100% unbiased -BPTrace halper use in helpers/vray

brute force on both primary & secondary = 100% analog of octane DIRECT KERNEL, maxdepth bounce=8. when i speed testing vray_vs_octane i use octane direct glossy=20 depth, refractive=20 depth 5000k samples, maxdepth=8 and vray 20x20+bf64-128samples at 8 maxdepth (secondary bounce). its 100% likely analog images generate.