What happened to DOF subdivisions under VrayPhysical Camera Settings?

[pixela]

Hi there,
Where is this setting moved?

[vlado]

Use the subdivs for the image sampler.

Best regards,
Vlado

[Dave_Wortley]

I found big differences when I did a test with a VRay 2.4 scene which had subdivision on the camera DOF and in Vray3, lots more noise just in the DOF blurring.

I had to increase the max samples from 5 to 48 and lower the noise threshold from 0.01 to 0.005 and color threshold from 0.05 to 0.005 to get the same quality again, amazingly it rendered 3 times faster and looks almost identical.

[rjohnson33]

I found big differences when I did a test with a VRay 2.4 scene which had subdivision on the camera DOF and in Vray3, lots more noise just in the DOF blurring.

I had to increase the max samples from 5 to 48 and lower the noise threshold from 0.01 to 0.005 and color threshold from 0.05 to 0.005 to get the same quality again, amazingly it rendered 3 times faster and looks almost identical.

That's interesting. Did you change any values in the material/light etc. subdivisions or just let those different threshold values clean everything up?

I had been wondering about handling DOF with regard to the recent optimisation tutorials -
so you take a scene and try and discern your max AA value right at the beginning with intended DOF in enabled (which is already a bit difficult to my mind but anyway).
You then try and take all of your material, light & GI subdivs to just over a noise free level to take the burden of the AA but you're now fixed to some degree to keeping your max AA value the same (because most of these individual subdivs are divisible by it).
What happens if you realise at the end that the max AA value isn't cutting it in terms of DOF or some textures? or you just have cameras with varying amounts of DOF in the same scene.

You have to raise the AA but do you balance this by upping the individual subdivs (in lights/materials etc.) - maybe using the global subdivision multiplier?
Or do you rely on the increased AA samples to take over in the cleaning up of noise and ignore the fact that it is reducing all the individuals subdivs? - I suppose this is the point of the max AA value being used to divide all other subdivs in the first instance.

Sounds like in your recent test that you're heading back to more of a 'Universal Setting' approach and maybe that makes more sense in a scenario where you know you're going to end up using heavy DOF (or any other factors that require higher AA sampling).

[Dave_Wortley]

No changes in material subdivs or lights. I tried to match the SamplerRate intensities as much as I could using the min/max and then just lowered the thresholds till the image went as clean.

[svetlozar.draganov]

I found big differences when I did a test with a VRay 2.4 scene which had subdivision on the camera DOF and in Vray3, lots more noise just in the DOF blurring.

I had to increase the max samples from 5 to 48 and lower the noise threshold from 0.01 to 0.005 and color threshold from 0.05 to 0.005 to get the same quality again, amazingly it rendered 3 times faster and looks almost identical.

It would be great if you could share with us 2.x vs 3.x final-result and render-times.

[svetlozar.draganov]

That's interesting. Did you change any values in the material/light etc. subdivisions or just let those different threshold values clean everything up?

I had been wondering about handling DOF with regard to the recent optimisation tutorials -
so you take a scene and try and discern your max AA value right at the beginning with intended DOF in enabled (which is already a bit difficult to my mind but anyway).
You then try and take all of your material, light & GI subdivs to just over a noise free level to take the burden of the AA but you're now fixed to some degree to keeping your max AA value the same (because most of these individual subdivs are divisible by it).
What happens if you realise at the end that the max AA value isn't cutting it in terms of DOF or some textures? or you just have cameras with varying amounts of DOF in the same scene.

You have to raise the AA but do you balance this by upping the individual subdivs (in lights/materials etc.) - maybe using the global subdivision multiplier?
Or do you rely on the increased AA samples to take over in the cleaning up of noise and ignore the fact that it is reducing all the individuals subdivs? - I suppose this is the point of the max AA value being used to divide all other subdivs in the first instance.

Sounds like in your recent test that you're heading back to more of a 'Universal Setting' approach and maybe that makes more sense in a scenario where you know you're going to end up using heavy DOF (or any other factors that require higher AA sampling).

In version 3.0 you can control the balance between Image Sampler and DMC Sampler via Min Shading Rate parameter:
http://docs.chaosgroup.com/pages/vie...MinShadingRate

You could even disable DMC rays division by Image sampler if you turn off Divide shading subdivs - please refer to the link above.

[Dave_Wortley]

It would be great if you could share with us 2.x vs 3.x final-result and render-times.

The scene has glossy refractions/reflections, HDRI, LightMtl, Caustics and DOF. Rendered at 2k with no noise.

Vray 2.40.04
AA 1/5
Clr Threshold 0.05
DMC Adaptive 0.95
DMC Noise 0.01

Render Time- 1:23:43

Same scene opened in VRay 3.0 with no changes, but more noise visible, mainly in the DOF.
Render Time: 0:52:02

Scene adjusted to try to match same shading rate and noise quality of V-Ray 2.4
AA 1/48
Clr Threshold 0.005
DMC Adaptive 0.95
DMC Noise 0.005

Render Time: 0:29:01

[svetlozar.draganov]

Nice to see those results
Thank you very much for your feedback.

[voltron7]

Nice results Dave. May I inquire about the other settings/differences between the two versions?
Others like Brute Force GI, camera DOF, light subs, etc.

Trying to get a sense of possible speed improvements from version 2-->3 and these are some of the more detailed (and good result at 2k) explanations!
thanks in advance, and no worries if you can't don't answer those