I thought it might be interesting to begin a new thread to discuss possible new approaches to sampling now that V-Ray 3.0 is out and provides two new (very powerful) controls over the samplers:
Min Shading Rate - Globally raises the Max Secondary Samples per primary sample to be no lower than this value. Divide Shading Samples - Enables or Disables the connection between the Image Sampler and DMC Sampler (AA Max Samples dividing DMC Samples).
I’ve added both of their functionality into the DMC Calculator for clarity as well.
I haven’t had the free time to do extensive testing, but I was thinking a new ‘full control’ method of working could be to:
- Disable the Divide Shading Samples checkbox and set all Lights/GI/Materials Subdivs values throughout a scene to 1.
- Set the Image Sampler (AA) settings according to the needs of the scene via the SampleRate element.
- Once the AA is set, bump up the various Subdiv values throughout the scene one at a time to systematically eliminate any sources of noise.
- This would give you full control of every sampling value in the scene - whatever you set is exactly what you’d get - in essence the polar opposite of the ‘Universal V-Ray Settings’
I’m sure there can be other novel approaches, so I’d love to hear any ideas and put some of these new methods to the test.
I would add, for those that have time and desire to go through everything and optimize it Some users get the wrong impression that you *have* to go through all that each and every time, whereas the reality is that often the time that it takes to do all that could be better spent while the computers are working. I myself often prefer the universal settings, following the “user’s time is more valuable” paradigm
Yes definitely! I totally understand the advantages and reasoning behind the Universal Settings - and whenever render time and CPU availability isn’t a huge concern, I tend not to bother with optimization either.
I just think it’s great that we have all the tools needed to do the job whichever way a user wants to approach it… whereas an unbiased engine like Arnold cant offer that level of control.
I know I’ve certainly had instances where I had to squeeze every last second out of the render times to make a tight deadline.
When you’re dropping frame times down from 3 hours to 40 minutes a few days optimizing doesn’t seem so bad compared to ‘miss the deadline because renders haven’t finished’
Yes, fairly sure.
‘Min Shading Rate’ raises the Max Secondary Samples per primary sample.
‘Min Samples’ raises the Min Secondary Samples per primary sample.
This times 100 :)… Really, I prefer my artists to focus on the look rather than spending too much time optimizing. That’s why I also love min shading rate… Combine it with universal and just focus on shaders/lighting.
From my last tests. Actually optimizing everything manually… Gives you just a small boost compared to universal in 3.0, by small I mean 10-15 % faster, sometimes even slower… And I still prefer nice constant noise (from universal) over everything in animation.
I know that sometimes it’s worth to optimize it a little further. I don’t deny that knowing how to do it is important. I just feel that for new users I would just recommend universal + optimizing through min shading rate. Makes it simple and produces amazing results.
By universal in 3.0, I mean universal + heavy usage of min shading rate.
I think that before 3.0, universal setup like this http://help.chaosgroup.com/vray/help/300R1/tutorials\_unisettings.htm was casting too many camera rays, and in 3.0 you can add min shading rate to force more secondary samples, this usually makes rendering a lot faster and scenes are extremely easy to setup.
I have to disagree about optimized scenes only being able to give a small boost (10%-15%) in render time compared to universal settings.
I’ve personally experienced certain production examples gain much larger boosts from optimization. But it obviously varies largely from scene to scene.
And even so - a 15% boost in render speed can certainly add up to a lot of saved time when dealing with hundreds / thousands of frames.
I really do like the idea of Universal Settings with adjusted Min Shanding Rate as a quick and easy compromise between the Universal and Full Control settings!
I know that in 2.0 the difference was huge, but in 3.0 it’s a bit different story. Is there any chance you could play with some of your optimized scenes vs 3.0 + min shading rate. Would love to see how it compares exactly.
From my experience so far it’s in range of 3-6 for scenes with DOF and a lot of details in texture and geometry. In architectural interiors with many flat surfaces it could be above 10 or 20…
It’s really easy to find a fastest shading rate for each scene, region or tile map in render mask works well.
Generally, yes. But don’t take it as a rule. Fastest shading rate will vary from scene to scene.
Also I wouldn’t call dof/mb heavy it just needs much more camera rays compared to usual architecture scenes. But overall it can be quite fast by actually lowering min shading rate (more AA/camera rays, less secondary samples/bf,lights,gloss).
By this do you mean that you’re taking a small test region and rendering out tests using different min. shading rates to see which one renders fastest?
(I guess a test area that covers worst case DOF, reflection and GI so you can make sure that the quality of these isn’t compromised by the setting)
Also, does anyone know why it’s mentioned in the help documentation that min. shading rate is particularly useful in conjunction with the progressive sampler? What would the relevant connection be between those two?