Glossy Reflections very slow on 3.3?

[FaustoDeMartini]

Not sure if anyone is seeing the same issue, but when I rendering a shader with a broad glossy refection using VRay Mtrl, I'm experiencing much higher render times compared to the previous version of VRay if I bump the Reflection subdivisions to 16 and 32.

I'm using the Adaptive Subdivision Image Sampler values of 2 (Max rate) and -1 (Min. Rate) which it was the same as I was using before.

[Morbid Angel]

why would you use adaptive subdivision?? its obsolete...

[vlado]

Not sure if anyone is seeing the same issue, but when I rendering a shader with a broad glossy refection using VRay Mtrl, I'm experiencing much higher render times compared to the previous version of VRay if I bump the Reflection subdivisions to 16 and 32.

Can you post a scene for this? I can't think of any changes that would cause this, but maybe there is something going on that I'm missing.

Best regards,
Vlado

[Alex_M]

Yes, I'm curious to know why you use Adaptive Subd sampler instead of Adaptive. I've never gotten it to render faster than Adaptive. Also if I remember correctly, Adaptive is much better than the Adaptive subd. sampler at rendering blurry/glossy effects such as very blurry material reflections, DOF, motion blur etc. Which your case seems to be. Detailed geometry, textures and fine detail in general are also a big problem for Adaptive Subd. AFAIK.

[FaustoDeMartini]

Thanks for the replies!

Dmitry- Not sure what you meant by Obsolete. I did a bunch of tests and each Sampler gives a different result. For me Adaptive Subdivision gives a good speed and less noise on the reflections... but maybe I'm missing some specific parameter to get less noisy reflections using Progressive and Adaptive.

Alex M- I have always used Adaptive Subdivision since it gave me a good balance between render time and less noisy renders. Please check the render tests I attached to see what I'm talking about But again, since I don't have much experience with the other samplers maybe I'm missing something. )

Vlado- Thanks for the fast reply, I will prepare the scene to save the area where I'm experiencing the issue.

Meanwhile, I attached three renders of the section I saw the slowdown. I also tested Adaptive, Progressive and Adaptive Subdivision. As you can the Adaptive SubD has the longest render time but the metal renders in a better quality.



Thanks!

[^Lele^]

Meanwhile, I attached three renders of the section I saw the slowdown. I also tested Adaptive, Progressive and Adaptive Subdivision. As you can the Adaptive SubD has the longest render time but the metal renders in a better quality.

That's expected: the ASubD sampler has to take the full amount of samples per sub-pixel, so it takes, with a max of 2, potentially (4^2) 16 times more samples than the other two samplers.

This behaviour was the one saving grace of the method for rendering beauties, imho, before 3.0, to make sure one had a firm hold on the sampling happening within the scene.
MSR changed that for 3.0, and SP3 changes it further still, making it the least appealing of the four AA choices, i dare say.
It still has excellent uses for data passes and such, but for the Beauty renders, i'd give a try to progressive and adaptive first.

With twice the rendertime (halve the noise threshold for progressive or adaptive) i'd wager you'd get a much closer image to ASubD.

[FaustoDeMartini]

That's expected: the ASubD sampler has to take the full amount of samples per sub-pixel, so it takes, with a max of 2, potentially (4^2) 16 times more samples than the other two samplers.

This behaviour was the one saving grace of the method for rendering beauties, imho, before 3.0, to make sure one had a firm hold on the sampling happening within the scene.
MSR changed that for 3.0, and SP3 changes it further still, making it the least appealing of the four AA choices, i dare say.
It still has excellent uses for data passes and such, but for the Beauty renders, i'd give a try to progressive and adaptive first.

With twice the rendertime (halve the noise threshold for progressive or adaptive) i'd wager you'd get a much closer image to ASubD.

Ya, it makes sense! With the tests I did, progressive seems the best balance between the slower Adaptive SubD render times and the much more Noisy Adaptive. Thanks a lot for the input!!!