Ready to watch this entire course?
Become a member and get unlimited access to the entire skills library of over 4,900 courses, including more CAD and personalized recommendations.Start Your Free Trial Now
- View Offline
- 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
Skill Level Intermediate
As the name suggests, the Adaptive DMC image sampling engine adds adaptivity into the image sampling process. This means that based on some user input parameters, the engine can make some very deliberate choices about where and how it places samples in our scenes. Again, to set up the Adaptive DMC engine, let's come into our Options dialog. Let's come into the Image Sampler rollout and this time, if we use the Type dropdown, we can see that Adaptive DMC is the second on the list.
The Adaptive DMC engine is indeed the default inside of V-Ray for SketchUp. So if you start a new scene, chances are you are going to be working with the Adaptive DMC engine straightaway. Now whilst we don't have an awful lot in the way of control parameters for this image sampling engine, clearly we have a little bit more than our fixed-rate engine. One set of parameters that control how this engine works are these Min and Max Subdiv values. These really are what give V-Ray the ability to play samples over a number of passes, passes that refine and optimize the sampling solution as they go.
The specific number of passes used will be controlled by the number of steps between the Min and Max values and this Color Threshold setting. In its first, or initial, pass V-Ray will always place the minimum number of samples, as set by the Minimum Subdivs parameters. Then using this Color Threshold value as a control mechanism, V-Ray can and will add more samples when needed, working up to either the Maximum Subdivs setting or indeed the cutoff point that has been determined by the Color Threshold value.
One note of warning here: we never want to set our Min and Max Subdiv rates to the same number. If we do that, we effectively kill the engine's adaptivity, and the DMC engine will simply function as per the fixed-rate engine. The renders we get will be identical. The render times that we get will be identical. Now the values we have here are the defaults for this particular engine type, but for production purposes, we generally want our defaults to give us fast feedback, particularly in the early stages of a project.
All we really need is a render that will give us a general idea of how the scene is looking. For this reason, we may want to actually save out some settings of our own from the Options Editor using lower values. So we could for instance set a Max Subdivs value of 3 and then we can come up to our Options Editor menu and just click on the Save icon, save those options out, and then whenever we want to work with fast test renders, we could just load those settings back in for ourselves. As the first of the renders we will examine inside of Photoshop use these particular settings, a Min value of 1 and a Max value of 3, we are going to stick with these for this moment in time.
Now before we look at our render, just to reiterate the process again, as we have now set the Max Subdivs value to 3, V-Ray will check the information it receives from the first, or initial, pass that it makes using the Min Subdivs value. It will measure that information against the threshold value, and if it decides that more samples are needed, then it will proceed on to a second pass, subdividing our pixels and placing more samples in the needed areas. And it will continue to do that until the maximum number of subdivisions is reached or again, until the Color Threshold cutoff is reached.
With these settings then, let's jump into Photoshop and have a look at our first render. As you can see, we get something very comparable to the initial renders from our fixed-rate engine. We get a pretty good idea of how our scene is progressing. Again, we can check things of just lighting, composition, and materials without any real trouble. We can make some good evaluations of these. We do still of course have a lot of noise in the scene, which again we are going to need to progressively clean up. And if we just have a look at our render times, so if I just use spacebar and if we just pan down, you can see we are slightly higher in terms of render times than our first fixed- rate render, which is not surprising really. Because we have so much noise that needs cleaning up in our scene, V-Ray at this moment in time is not using our minimum subdivision setting.
Clearly we are using the three subdivs that we've set in our Max parameter. Hence the slightly cleaner render that we have and the increased render time. Now if we just zoom in to have a look at our materials, you can see our lines on this particular sphere are looking much cleaner than our initial fixed-rate render, so we can tell most sampling is going on there. And we can see just generally that the materials are looking a little bit better, although the diagonal lines on our lead sphere here are not looking too good at this moment in time.
Let's go back into SketchUp and see what we can do to improve this situation. To do this, we are naturally going to want to increase our maximum Subdivs parameters. Now you may be wondering why we are not increasing the minimum subdivisions. Typically speaking, we want to leave this as low as possible, in order to give the engine the ability to be adaptive. The general recommendation is that we have at least three steps in between the values set in the Min and Max Subdivs setting. For this demonstration, however, we are just going to keep our Minimum Subdivs set to 1 as we increase our Maximum Subdivs value.
With 1 and 8 set then, let's jump into Photoshop and let's see what we get. Here we go from our 1 and 3 render to our 1 and 8, and you can see, things do, as we would expect, clean up quite considerably. The noise on our wall cleans up very, very nicely indeed. The reflections in our materials also improve quite considerably, as do the lines on our lead sphere here. And again watch for the small noise in our materials; you can see that we definitely pick up a lot more detail there also. And our shadow edges clean up nicely.
We can see that that there is a lot less noise, a lot less grain contained inside those as well. As you would expect though, the improvement in quality comes at a cost, so we are now at 8 and a quarter minutes, which is quite high. We've made quite a jump in terms of render time percentage, just as we did of course with our fixed-rate engine. Now again, because of the noise inside of this scene, we are not really getting a chance for V-Ray to do any adaptive sampling. This scene really is not the best one in terms of showing off how the adaptivity works.
What we are seeing though is just how well the image sampling engines can handle a difficult scenario. Lots and lots of noise, fine-line detail, very fine specular detail in some of our materials, and yet our image sampling engine can handle that nicely. Well, again let's increase the sampling quality. Let's go from 1 and 8 to 1 and 16 now, and we would go from this render to this, which again, as you can see, doesn't affect every area of the scene quite so obviously. The noise on our wall clearly cleans up quite considerably, but our materials don't seem to do an awful lot.
Although if we just zoom in, and again if we go to our 1 and 16 and use Ctrl+Plus on the keyboard to zoom in, you will see that we are definitely getting cleanup in the reflections. We are losing some off the noise that we still have in here, and if you keep an eye on the fine noise detail here, you'll see that we do pick up quite a bit more detail with our 1-and-16 render. So we are definitely making a difference. And of course, those are reflected in our render times. We are now at 27 and a quarter minutes. Although, if you remember, when we used our fixed-rate engine, when we had 60 subdivisions there, we were at around about 32 minutes.
So for comparable quality, we are actually dropping down the render times now, as V-Ray as it reaches the limit of what it needs in terms of samples to clean up the scene, as V-Ray now is able to actually bring in a little bit of that adaptivity. And that of course is making a bit of difference to the render times now. If we just jump back into SketchUp, one of the things we of course have not explored up until this point is the Color Threshold value. Well, let's make a change now. Now that we have such a high setting--and remember, we are actually working with not 1 and 8, but 1 and 16 at this moment in time.
Now that we have those high settings in there, because remember, with a maximum Subdiv of 16 set, we are actually working with 256 samples, or rays, per pixel. Well, with that setting, let's go and make change to our Color Threshold value. Let's set that to 0.003, and again, let's jump back into Photoshop to see what that would do. This means we would go from our standard 116 to our Color Threshold version. And on initial comparison, there doesn't seem to be a lot of difference.
Both of these images look to be producing pretty much the same result, although again, if we just zoom in on each of these images, you will be able to see that we are using extra samples. If you just keep an eye on the reflections in our lead sphere here, you can see that they do clean up quite considerably, really. There is quite a bit of fine noise detail in there, and those extra samples that are now being allowed to work by the Color Threshold setting are actually cleaning that up very nicely. And if we just zoom out, you can see that we are indeed using more samples in the scene, because our render time has gone up by a number of minutes, which again demonstrates to us that by lowering the Color Threshold value we are allowing V-Ray to make more use of that Maximum Subdivision setting.
So the adaptivity then clearly at work, even in this difficult-to-render scene. And as we say, this really is not the best of scenes to show off the quality of the adaptivity in this engine. Really, what we would want is a scene that has lots of clean areas to it, maybe just straight block colors without lots of noise. Then you would see quite a difference in terms of the render times between the Adaptive DMC engine and the fixed-rate one. But we do have adaptivity working. We can see that a little bit, even in this difficult scenario. And adaptivity really is what makes this engine, in my humble opinion, generally speaking, the best choice as a sampling engine for most production situations.
One extra benefit of the ability of this render engine to work adaptively is that it gives excellent results in terms of quality and render speed whenever there are blurry effects in our scenes. By that we mean whenever global illumination, depth of field, motion blur, or blurry reflections and refractions are at work. On top of that, the fact that the DMC sampler doesn't need to hold sample information in memory can be a huge benefit, particularly if we are rendering scenes that have high memory requirements.
SketchUp of course being a 32-bit application means that use of memory is at a premium at all times, so just that simple fact alone can make it an excellent choice just to be able to get our scenes rendered. Incidentally, the same is also true of the fixed-rate engine. It too doesn't hold sample information in memory. Now in terms of weaknesses, there really aren't any as such that I would ascribe to the Adaptive DMC engine. It is capable of producing the very highest quality, but it can also be configured to work a little bit more speedily for us if that is what our current project needs.
In fact, the only thing that we could say against this particular engine is that whilst it is clearly capable of being adaptive in its sampling approach, it is not able to perform any kind of undersampling or infrasampling. In our next video we will take a look at of course the last of V-Ray's image sampling engines, one that is indeed capable of performing under- or infrasampling, and this is the adaptive subdivision engine.