Join Brian Bradley for an in-depth discussion in this video The bucket sampler, part of SketchUp: Rendering with V-Ray 3.
- [Instructor] The adaptive or bucket sampling system in V-Ray adds controllable adaptivity to the image sampling process meaning that based on a number of user-specified input parameters the engine can and will make some very deliberate choices about where and how it decides to use the sample counts that have been made available to it. To enable this system in V-Ray for SketchUp all we need do really is turn the default progressive option off and when we hit the render button, we see that once the light cache precalc is finished, we now render not with a series of continuous refinements but with a bucket system that methodically renders through the various areas of the image bringing them up to the specified quality level.
The controls for this system are similar to those found on the progressive sampler only this time along with the quality slider and noise limit value we get these min and max subdiv options. Now, right out of the gate here we do need to note that we would never really want to be setting the same value in both the min and max fields as doing so would kill adaptivity in the system and so cause our render times to be unnecessarily high. To show how the different settings that we can use here will affect our renders, I have saved a number of pre-rendered images to the history list so that we can use them to compare a variety of min and max subdiv settings.
Changing nothing else about the sampling defaults then, min and max subdiv values of one and four as seen in our first image give us a pretty decent representation of lighting and materials in the scene. If we push in on geometry edges, we can see that things are perhaps a little ropy in areas and of course, there is plenty of noise that needs cleaning up in the image. Still, this one K render took just five minutes to complete on this pretty low end I74820K CPU running at 3.6 gigahertz.
Let's try to clean things up a bit though by pushing the min and max subdivs to one and 12. What we see as we load our second render into the frame buffer is that whilst we have obviously started to clean up noise in the image, there is still plenty there. Our geometry edges whilst looking better still have a way to go and we have of course considerably increased our render times. Well, let's make a huge jump in the number of samples being used in the scene by setting min and max values of one and 50 then.
As we load our third image into the frame buffer window what we see now may actually surprise us even more as first of all our render time has stayed pretty much the same which probably clues us that as we zoom in, we also see that the image itself hasn't really changed all that much. Indeed, one thing we need to realize about the adaptive or bucket sampler is that it doesn't automatically make use of all the samples made available through the max subdivs parameter because certain areas of our image may not meet the criteria for all of those samples to be used, that criteria being determined by the noise limit value found in the raytrace rollout.
Let's see what happens then when we drop that value from its default of 0.05 to a high-quality production setting of 0.005. Loading our fourth image into the frame buffer and comparing it to our previous min one, max 50 render, we see that we have now taken quite a big leap both in terms of render times and in terms of image quality with noise levels in the image being more than acceptable at this point. The render times have jumped so much because with our current setup, we are letting the primary samples in V-Ray do most of the work and unfortunately when it comes to image sampling, primary rays are the most computationally expensive around.
One way to improve the situation would be to tell V-Ray to let secondary samples in the scene take more of the load. An easy way to do this using just a single global parameter is by means of the shading rate option found in the optimizations rollout of the advanced UI. Higher shading rate values tell V-Ray to spend less time using the primary rays and instead focus more effort on sampling using the secondary rays for noisy effects such as glossy reflections, GI, area shadows and so on.
With our min and max subdivs set back down to one and 12 then, let's set our shading rate up to a really high value of 24 and take a look at what that would do in our fifth and final render which although as you can see in a comparison makes no discernible difference to the final quality of the render, even when zoomed in, does shave quite a bit off the overall render time which generally speaking is always going to be a welcome scenario.
- Gamma handling in V-Ray 3 for SketchUp
- Working with interactive rendering
- V-Ray light types
- Working with irradiance mapping
- Rendering animations
- Working with the V-Ray camera
- Using the Materials UI
- V-Ray FX tools
- Stereoscopic 3D rendering
- Using V-Ray objects