6 Ways You May Be Sabotaging Your Blender Render Settings
Depending on your workstation’s specifications, rendering a scene can be an excruciatingly slow process. If upgrading your hardware isn’t an option, you may try to change your settings to speed up rendering, which often sacrifices quality in the process. Let’s take a look at some things to watch out for and what you can do instead!
Rendering locally means using your own workstation to render scenes. Many new and seasoned designers alike do this for test renders and animations, but rarely does a final render finish rendering as quickly as you want it to.
2K/4K renders and AOV passes are very processor-heavy operations. If you’re running your 3D software or renderer on a weak system, the first thing you’re likely to end up doing is dropping your sample count and that alone will drastically decrease the quality of your final render.
An easy solution to this problem is to move your local rendering to a cloud-based service. At Render Pool, our affordable render farm can greatly increase the quality of your Blender renders at the speed you want.
Limiting Max Light Bounces
The backbone of any successful rendering engine is its ability to accurately compute light and how it bounces on objects. This is known as ray tracing. One mode of thought comes to mind: “What is the bounce setting’s influence on noise in renders? I’ve been told that if I want less noise in my image, I should reduce the bounces.”
The problem with this thinking is that it defeats the purpose of having any light paths or bounces at all. In the real physical world, there are an infinite number of light bounces. The only way one can try to simulate those bounces is to have a lot of them, or else, the shadows will certainly be darker than a similar scene in the real world.
Blender’s default Max Bounces count is 12. In practice, a slightly smaller number of bounces is enough to simulate real-world light patterns. With that in mind, you can set a maximum value of four to six bounces, but anything below that will result in a significant decline in quality in your final renders.
Low Clamping Values
Another area of debate in the Light Paths section is clamping. What clamping does is limit the high brightness values of a scene and help reduce fireflies (dead, burnt-out pixels). This value is typically set to cut out-of-range or garishly bright samples in a scene, but some users tend to set their clamp settings too low thinking that this will save on light bounces and therefore improve render times.
The only issue with this is that if the clamp values are set too low, the image quality will blur and darken, giving your render a thin, low-contrast look. When you change clamp values, be sure to pay close attention to the brightest samples of your scene when making changes.
The best solution, though tricky, is to set the clamp value low enough to help reduce noise, but not so much as to reduce the quality of the final image. One work-around to overcome this is to allow only the brightest values to darken, and then to correct it later in post-processing with Adobe Photoshop or After Effects.
Material work-arounds, or “hacks” for surfaces with reflections or other complicated elements, are quite popular in the 3D community as they presumably keep render times low without sacrificing quality.
For the third-party Radeon ProRender engine, materials with reflective or refractive properties are readily available in its RPR material library. However, there aren’t many work-arounds for Blender’s Cycles engine as it’s an unbiased ray tracing engine, meaning it’s physically accurate when processing the behaviors of light.
The work-arounds that are available for Cycles… don’t look very good. Moreover, they often end up affecting the visual quality of your final render. Therefore, it’s recommended to avoid using material hacks or solutions that use glossy or transparency shaders, because unbiased rendering engines don’t give the same results when compared to biased engines like Eevee that are actually effective at faking the behaviors of light while still maintaining speed.
Object Mesh Optimizations
Object mesh optimizations refer to keeping polygon counts down by using low levels in your subdivision and decimation modifiers. The idea is that a low-poly mesh will render faster; and while this is true to some degree, it has little to do with rendering and is much more about memory. Dense meshes don’t have an effect on the CPU or GPU when rendering a Cycles tile. Instead, the bounding volume hierarchy (BVH) process is what takes the hit.
Bounding volume hierarchies are used to efficiently support several operations on sets of geometric objects, such as ray tracing acceleration. During rendering, this process relies exclusively on how much memory you have. If you’re running on a machine with low memory (4 or 8 GB, for example), that is what will make your render take a long time. While the optimal minimum of RAM you should have when working in 3D software is 16 GBs, you can render scenes at lower capacities as well. It will take a little longer, but as long as you don’t have extremely dense meshes in massively large environments, over-optimizing your object’s poly count is unnecessary.
Too Much Denoising
Denoising is often considered the last attempt to remove noise from an image before publishing. There are quite a few programs that specialize in denoising, both third-party and internal, as well as post-processing stand-alone programs like Neat Video.
Radeon ProRender for Blender arguably has the best solution as its new and recently improved RPR2 plugin has made manual denoising setup completely obsolete, but for Cycles’ users, it’s still a regular practice. However, no matter what method you use for denoising, the most important thing is using it in moderation.
Too much denoising can create visual artifacts that are left behind in darker areas due to most denoising and smoothing algorithms. These artifacts can be glaringly obvious and can potentially reduce the quality of your final render. You have to find a comfortable medium between the noise and the smoothing artifacts brought on by these algorithms. It takes a bit of time to find that perfect harmony, but as long as you’re using this technique in moderation, chances are you won’t have a problem.