Join Brian Bradley for an in-depth discussion in this video Working with the adaptive subdivision engine, part of V-Ray 3.0 for 3ds Max Essential Training.
- The adaptive subdivision engine that we will look at in this exercise works a little differently as compared to the other three options available in that, instead of creating an internal pixel array that becomes more and more subdivided as the "Subdiv" parameter is increased, this engine instead creates a grid that is set to the same resolution or size as our render output and that then acts as an internal overlay. This is used by V-Ray to position or place samples in the seed. After making a first pass, the samples are compared and if the difference between any two samples is greater than the value set in the color threshold option, then the grid will be subdivided and more samples added.
We do, if I just open up the "Render Setup" dialogue, have a couple of super-sampling options available. We have "Object outline" which will over-sample the outline of geometric objects in the scene. We have a "Normals threshold" option, which will super-sample objects whose normals have widely varying directions. And we also have a "Randomize samples" option, which is checked by default. This helps the system produce better quality anti-aliasing with regard to any horizontal or vertical lines that may be found in the scene. Now you may already have noticed that instead of min and max subdivision values here, we have a min and max rate setting.
These rate controls work quite a bit differently than the straightforward subdivision method used by the other engines and so we do need to be a little bit careful when setting values in here. To understand why I say that, with a "Max. rate" of 0 set, we can use just a single sample per pixel. If we increase this by just a single step, however, so going up to a value of 1, we could now be using a maximum of four samples per pixel. If we jump to a value of 3 here, we could be using 64 samples per pixel, and as we continue to increase the rate value, that kind of exponential growth in samples continues.
By the time we are using a "Max. rate" of around 5 or 6 then, we could actually be pumping huge amounts of samples per pixel into our scene, which could most definitely slow our system down. Let's open up the "V-Ray frame buffer" window then, load the first image from the history window, and take a look at what the scene would look like if rendered using the current adaptive subdivision defaults. As we would perhaps expect, things are a little rough with lots of shadow noise and missing detail on our textures.
Interestingly, the render times aren't even particularly fast here, coming in at just short of three minutes. Well, what happens if we start to increase the "Max. rate" value? Keeping in mind of course that the current default rate of 2 gives us a possible maximum of 16 samples per pixel. Well let's make the jump to 3, giving us a possible total of 64 samples per pixel, and load our next image into the "V-Ray frame buffer", setting our original to show up in channel A. Well, whilst we can see that render times do go up by a small amount, the quality doesn't actually appear to change at all.
This is because, as with the adaptive engine, the threshold value, this time "Color threshold", can make a big difference regarding how many of the max samples available, the engine actually makes use of. Let's set the thershold here to a more sensitive value of 0.05 and load our third render. What we see is that render times now start to climb by quite a bit, and we see the quality does start to improve with this engine being very good at getting nice, clean edges around our geometry, even without the "Object outline" super-sampler enabled.
The problems seen in our initial render are clearly still with us, as we are missing quite a bit of detail from our jar textures. The engine's weakness when working with blurry reflections having quite an impact here. Clearly then, we need to be more aggressive in our use of samples. So let's drop our threshold value down to 0.01, and take a look at what we get. Which is a much cleaner end result, with both shadow noise and detail textures looking much more acceptable. We have, of course, taken a very big hit now in terms of render times, coming in just short of the 30 minute mark here.
And of course, one thing we would also have to keep in mind here is the fact that the global DMC sample controls will still affect any noisy or blurry render effects that we have in our scene. And so we may need to work with those controls as well in order to clean up some areas of our image. Now in terms of strengths, the adaptive subdivision system works really well in scenes that have large areas of block color. And so if we are working with that kind of a scene setup, then this engine may well be able to produce renders very quickly indeed for us.
Of course, there are also weaknesses. One is the fact that the adaptive subdivision engine is extremely memory intensive, as all of the sampling information gets held in memory. So unless we have a really well spec'd machine in terms of ram, we could find ourselves into "out of memory" errors. We have also seen here, that this engine is not particularly good at working quickly with scenes that have lots of noisy effects such as blurred reflections and area shadows. If we choose to play to its strengths though, the adaptive subdivision engine can be a very powerful production tool.
Next up, we will take a look at the newest of V-Ray's four image sampling engines, and one that can provide certain artists and workflows with big productivity boosts, this being the progressive image sampler.
- Using the new UI elements, Quick Settings, and revamped Frame Buffer
- Understanding color mapping modes
- Adding V-Ray light types
- Working with the V-Ray Sun and Sky systems and dome light
- Using irradiance mapping and light cache
- Working with diffuse color maps
- Making reflective materials
- Creating a translucency effect
- Using the new SSS and skin shaders
- Ensuring quality with image sampling
- Working with the adaptive subdivision engine
- Controlling the physical camera
- Working with FX tools such as VRayFur and VRayMetaball
- Stereoscopic 3D rendering
- Using Render Mask
Skill Level Intermediate
Q: This course was updated on 02/02/2016. What changed?
A: We added tutorials on the new 3ds Max camera tool, which replaces the defunct V-Ray Physical Camera. The author also includes a method for creating a V-Ray camera via scripting.
Q: This course was updated on 04/19/2018. What changed?
A: New videos were added that cover V-Ray 3.1 to 3.3 updates.