Controlling the Adaptive Subdivision sampler
Video: Controlling the Adaptive Subdivision samplerThe Adaptive Subdivision engine is an extremely powerful image sampler that in certain circumstances has the ability to save us lots of render time while still delivering lots of render quality. To examine its control set, we of course need to go into our Options Editor, into the Image Sampler rollout, and in the Type dropdown we can set the Adaptive Subdivision as our image sampling engine. Now this particular engine works little differently than the Fixed Rate and DMC engines in that instead of creating and subdividing a pixel array internally, the Adaptive Subdivision engine instead creates a grid, set to the same resolution or size as our rendered output that it uses to position or place samples in the scene.
- What's next?
Viewers: in countries Watching now:
Create highly realistic 3D architectural drawings with V-Ray, a popular third-party renderer for SketchUp. This course shows how to take a single scene with interior/exterior elements and add lights, move cameras, and enhance objects with translucent and reflective surfaces. Author Brian Bradley explains concepts like irradiance mapping, perspective correction, and fixed rate sampling, while showing how to leverage each of the V-Ray tools and its material and lighting types to achieve specific effects.
- Installing V-Ray
- Creating natural daylight with V-Ray Sun and Sky
- Bouncing light around with irradiance mapping and light caches
- Setting up a depth-of-field effect
- Creating diffuse and reflective surfaces
- Working with the Adaptive DMC engine
- Manipulating color mapping
- Adding caustic lighting and occlusion effects
Controlling the Adaptive Subdivision sampler
The Adaptive Subdivision engine is an extremely powerful image sampler that in certain circumstances has the ability to save us lots of render time while still delivering lots of render quality. To examine its control set, we of course need to go into our Options Editor, into the Image Sampler rollout, and in the Type dropdown we can set the Adaptive Subdivision as our image sampling engine. Now this particular engine works little differently than the Fixed Rate and DMC engines in that instead of creating and subdividing a pixel array internally, the Adaptive Subdivision engine instead creates a grid, set to the same resolution or size as our rendered output that it uses to position or place samples in the scene.
Again, this all takes place internally. After making a first pass and placing samples according to the Minimum rate setting, a comparison is made and if the difference between any two samples is greater than the value set by our Threshold control, then the grid will be subdivided and more samples will be added as required. The Min and Max Rate values also work a little differently than the Min and Max Subdiv settings inside of the Adaptive DMC engine.
In fact, if we are not careful with these controls, we can very easily bring our system to a crawl. You see, when we are using a Rate value of 0 in this particular engine, we are actually using a single sample per pixel to gather information from over scene. If we set a Rate value of 1, we are now using four samples per pixel. But by the time we move up to using a Rate value of 3 in the Adaptive Subdivision engine, which of course would be 9 samples per pixel in the Fixed Rate and Adaptive DMC engines, we would actually be working with 64 samples per pixel. And that kind of exponential increase in the number of samples used continues as we increase our Rate value, just even by a single step.
Hopefully, you can see then that we cannot use the same kind of Rate values inside of the Adaptive Subdivision engine as we would use in the Min and Max Subdivs inside of the Adaptive DMC system. If we did, we would possibly bring our system to its knees, simply because of the sheer number of samples that we would be asking it to compute. Now, the values that we currently have inside of our Adaptive Subdivision controls are indeed the defaults inside of V-Ray for SketchUp. And again, as with the Adaptive DMC engine, for starting out with test renders inside of a project, these are little high for my tastes.
So for the purpose of looking at over test renders, we're just going to make a little tweak to these settings. We will leave our Minimum rate at a value of -2. This means that we can undersample areas of the image if our image sampling engine determines that that is acceptable. With a Rate of -2, we are using just one single sample for 4 pixels in our rendered output. What we really need to change is our Max Rate setting. This we want to drop down all the way to 0. This means we are just using one single sample per pixel for our Maximum Rate.
So leaving our Threshold value at its default setting, let's jump into Photoshop and have a look at the render that these settings would give us. And as you've perhaps come to expect by now, you can see that over initial render is extremely low quality. It is extremely noisy indeed. But again, as with our previous engine types, we can still make a fair evaluation of what is going on in the scene. We can very readily tell what's happening with our lighting, our GI, and to some extent, our materials. What we do of course want from such low quality are fast render times, so let's choose our spacebar and then left- mouse-click to pan down and have a look, and you can see we getting this render at just over a minute and a half, which for a 1280 x 720 render, is very fast indeed.
So with those initial settings, yes, we are getting low quality, but of course we are getting very fast feedback, which is perfect for the start of a project. Naturally though, we're going to want to step up the quality level a little bit. We're going to want to see what we can get if we just increase our Max Rate settings a little bit. In fact, we will set our Max Rate up to value of 2, and we would go from this to this, which as you can see, is a reasonable jump in quality. The noise on our wall cleans up quite nicely. We start to see much more detail in our noisy materials, and of course the reflections start to take shape very nicely indeed.
With the Adaptive Subdivision engine, however, the problem can come in terms of render time. So let's check what we get here, and you can see, now we've jumped all the way up to just over 5 and a quarter minute. Now that may be acceptable, because we have taken a reasonable step forward in terms of quality and still, five and quarter minutes is not a huge amount of time just to get a nice reasonable-quality test render back. Again, if we just zoom out and back in to recenter that, of course, what we have at this moment in time is not acceptable as final render quality.
We would need to step things up a little bit more than this before we could say we were happy to show this to a client as a proposed final piece of work. So how would we do that in the Adaptive Subdivision engine? Well, our first thought would probably be to increase over Max Rate value, but we've already mentioned how dangerous that can be in terms of the system resources that we can eat up. The answer to improved quality in fact lies not with increasing our Max Rate value at each step, but rather, looking at balancing out our maximum rate setting with our Threshold value.
So let's jump back into SketchUp and make a tweak to that parameter. What we will do then is instead of accepting our default setting of 0.15, we will drop our Threshold setting all the way down to .01. With that change, let's jump back in to Photoshop and see what a difference that has made. Now remember, as we make the switch between these two images, we have not increased our Max Rate setting at all. What happened now is we would go from this to this, which as you can see, is indeed quite a dramatic cleanup.
The noise on the walls cleans up very nicely indeed. We get lots more detail from our noisy materials. You can see lots of the gaps that were there in our previous render are all now filled in with some very fine detail, and of course our reflections are looking much cleaner. Now you can see what a difference that Threshold setting has made to the number of samples being used when we examine the render time for this image. So remember, we were at just over five minutes with our previous render. In this instance, we are all the way up to 22.5 minutes, which is quite a considerable jump.
Clearly, you can see how sensitive the parameters are inside of the Adaptive Subdivision engine and why we need to be very careful with the values that we are using in there. Well, let's just recenter our image, because we have one more quality jump that we want to make. We want to increase now our Max Rate setting up to a value of three. We're keeping our Threshold 8.01 and if we do that, we would go from this to this, which if you keep an eye on the noise that is appearing on our back wall, you can see, does make quite a difference.
We really do clean that up very nicely indeed. However, our materials may appear not to really increase in terms of quality too much. But if we just use Ctrl+Plus on the keyboard to zoom in, just keep your eye on these reflections. And let's again switch our image, and we do of course want to zoom in and then just make the comparison. And you can clearly see that the noise does indeed clean up very nicely. Of course, again render times are going to be important to us, so let's zoom back out on that particular image and take a look at the render time, and you'll see that now we're all the way up to an hour on 60 minutes, which is quite a render time, which if we're being honest, was wholly expected for this particular scene.
The Adaptive Subdivision engine was never going to perform too well in this particular environment. As we mentioned, we have forced lots and lots of noise into this particular render, and that is not something that the Adaptive Subdivision engine copes with very well, especially not in conjunction with lots of blurry reflections, which we do have on our materials and our floor, and of course we have noise coming from our area shadows as well. The Adaptive Subdivision engine really functions well in scenes that have lots of flat color.
In those situations, it can make use of its undersampling capabilities. This means it can give is both quality and speed in our final renders. Whilst render times may clearly be an issue in certain instances when we use the Adaptive Subdivision engine, there is another weakness that we need to point out in connection with it. This is the fact that the Adaptive Subdivision engine needs to hold all of its sampling information in memory. This means that if we have scenes that already have high memory requirements, well, we could find ourselves running into out-of- memory crashes very, very quickly.
In this chapter then, we have examined all three of V-Ray's image sampling engines. Hopefully, we have demonstrated that they all capable of producing high-quality renders. Of course, there are differences in the way that each of these engine types are working that may make them suited to one particular project whilst not suitable for another form. And of course the choice of which one you use is entirely up to you. Project requirements, the time you have available, maybe even artistic preferences, all of these will play a part regarding the choice, regarding the settings that you use when you working with V-Ray's image samplers.
Hopefully though, the overview we've given in this chapter will be able to serve as a nice strong foundation from which you can build you own knowledge and understanding of V-Ray's image sampling engines.
There are currently no FAQs about SketchUp Rendering Using V-Ray.