V-RAY FOR SOFTIMAGE BETA - finaly here

[McNistor]

I believe min-16 is a bit too high. You have there a lot of "flat" areas where oversampling is wasted.
Try
min = 1-4
max = 64-128
cutoff = .1-.2
quality = go above 1
Going above 1 might increase render time (which I hope will be compensated by getting this to be more adaptive by lowering min samples) but it should help a lot in remove noise in DoF, reflections, etc.
See what happens and let us know.

[gustavoeb]

hey zafar would you care to share the scene?

thanks

[Zafar Iqbal]

McNister: Just been trying and I'm not getting decent results. Lowering the cuttoff doesn't seem to do anything/much and it's only when I increase quality a lot that I can get decent quality.

I'm currently rendering with 4, 128, 10 and 0,1 settings just for the sake of curiosity.

gustavoeb: I'd love to but I cant. It's not mine to share, sorry

[McNistor]

10 for quality is way too high I think. Anyway, we're waiting...

[gustavoeb]

I really like the higher contrast, and deeper gray of the MR renderings.

You may even like it. But it is not phisically-correct. It actually looks like the GI trace depth has been dialed down.

[Zafar Iqbal]

mental ray, 7h 47mins
- min 4
- max 128
- quality 15
- cutoff 0,01
Triangle filter @ 2
Brute force final gather @ 128 rays. 3 bounces.
Bookeh shader with 1 samples
1 real model and 9 instances.



Did another that took 8h 26mins
- min 2
- max 512
- quality 10
- cutoff 0,1
Triangle filter @ 2
Brute force final gather @ 128 rays. 3 bounces.
Bookeh shader with 10 samples
1 real model and 9 instances.

It looked slightly worse.

And again, the old VRay render again for convenience (3h 10 mins)


I do not think I can get get the mr render time much lower. The only thing left as for this comparison goes, is to do a render using Adaptive Sampling.

[Zafar Iqbal]

I've prepped a couple of scenes for anyone interested to give a try - read further below.

New render with mental ray using Adaptive Sampling
- min 0
- max 2
- contrast 0,05
Triangle filter @ 2
Brute force final gather @ 128 rays. 3 bounces.
Bookeh shader with 8 samples
1 real model and 9 instances.

6 hours



It's not happening - the only good render time I had was when FG samples were reduced to 8, however that affected GI precision, so that's not an option. 128 samples is the lowest I'm willing to go.

I also tried with Irradiance Particles, but it doesn't support geo "lights" so I dropped it again. I'll prep a scene with more exciting lighting and do some testing soon.

---

I've prepped some scenes - they both originate from the walker scene. Not that it matters too much but they are set up for 3delight, mental ray and VRay and produce somewhat equal look regarding light and materials, however do let me know if I did something terribly wrong.

One with cubes - yes, just as thrilling as it sounds but it's intensive and can put the CPU to some good use.

The second one contains the epic XYZ-RGB Dragon model containing 7.2 million triangles. 3dl and mr uses 2 additional instances while VRay has proxies for it.

Camera in both scenes is animated for GI flicker testing. Cubes are animated as well - nothing fancy though.

Donwload link (240MB): http://www.megaupload.com/?d=JZ5I3HGY

Please note that I've purposely stayed away from any kind of interpolation in my testing so far. If you are going to give it a shot yourself to speed things up, then please keep this in mind. The quality I'm personally aiming for is for animation but with post-production in mind - meaning somewhere between "typical" animation and stills.

Previews:

[mosy252]

please somebody tell me how can i render linear with 2.2 gamma in vray softimage ?

[Zafar Iqbal]

Disable your gamma in XSI if it's enabled and then add 2.2 Gamma in VRay's render settings under Color Mapping.

[gustavoeb]

I would say if you are rendering to EXRs you can keep to the way you render linear with MentalRay, no problem. Thats because they are float files and float files should NOT be saved with the gamma baked in.
If you want to save 8 bit files then you should bake the gamma in and sadly VRay does not support this checkerbox in Softimage:

(I'll try posting this at Vray's Forum)


In case you render EXRs and stick to the regular linear workflow in Si, there is something really important to do in order to reduce noise in Vray. You can tell vray the image is sRGB so it optimizes its sampling for this situation. This is how you do it:



And of course like in the original Si workflow don't forget that your image nodes should be set correctly (linear for linear images and sRGB for images with the gamma baked in).

Cheers

[Bellsey]

I've started to do some Vray Softimage tests as well. I've chosen a simple image and have started with the Universal Vray Settings:

GI: on; Primary Engine - Brute Force; Secondary Engine - Light Cache
Image Sampler: Adaptive DMC; Min SubD - 1, Max 100
DMC Sampler: Adaptive amount - 1; Noise Threshold 0.005
The image was rendered at HD 1280 x 720 and has one VRay Sun Light in the scene: Intensity - 0.008; Size Multi 20, Shadow Subdiv 16, Bias 0.02
Everything else is the default.
Render time was 0h 19m 15.7s - HP Elitebook 8730w, Quad Core, 8GB Ram

VR_test_001.jpg

[mosy252]

thank u guys.
u mean apply a color correction to texture with 0.4545 gamma?
so i cant render 8 bit images with SRGB like in max?right?

[Zafar Iqbal]

thank u guys.
u mean apply a color correction to texture with 0.4545 gamma?
so i cant render 8 bit images with SRGB like in max?right?

Sure you can, and XSI is very good at this - make sure the color profile in the tekstures property is set to sRGB. no need for any color correction nodes.

This might help: http://kobayashystips.blogspot.com/2009 ... image.html

[gustavoeb]

thanks zafar, that is what I meant.
Just make sure you have it on. But SI is pretty good about that, you dont need any extra bells and whisles. Its all in the docs

[mosy252]

i got it.thank u tahnk u tahnk u very much.

[Zafar Iqbal]

Graham: Did you use exposure control? I'm wondering how you got color bleeding like that. I can't seem to get any unless I use exposure controls/tonemapping.

I thought I already posted the following - apparantly not:

mental ray, 14 mins 03 secs
Adaptive Sampling
- min 0
- max 2
- Sampling contrast 0,05

Lightbox/Area Light
- samples 12, 12

Triangle filter @ 2
3 duplicated Dragon meshes.




VRay, 24 mins
Adaptive subdivision
- min 0
- max 2
- threshold 0,05

DMC Sampler
- Adaptive amount 1
- Noise thres. 0

Plane lights
- sampling 32 subdivs

Area filter @ 1,5
3 VRMeshes



Noise level is very similar IMO.

mental ray's actual render time was 1½ mins shorter. Rest went to "translation" whereas VRay needed much much less time (+1 to Vray). I couldn't use instances because you can't apply different materials to instances (I can with VRMeshes. +1 to VRay again even though they aren't the equivalents to instances).

The light to the right covers a bit in the VRay render. This means without it (didn't know how to make it invisible) VRay would have been slower by perhaps another minute or so.

Ps. for some reason my meshes aren't perfectly aligned any more. they were when I made the scenes that I uploaded, but later the VRMeshes were offset - in the viewports, that is. hm...

Update:
Exact same settings as before but with real geometry instead of VRMeshes

VRay, 13 mins 2 sec