I’m looking to start some projects, but render times could be an issue. So, I want to limit my scenes so that render times are manageable for me. What things in a scene contribute greatly to render times?
I know about noise filtering and tweaking the render settings; I’m looking for anything regarding scene creation. I expect that most things affect it to a degree, but I don’t really know how or to what degree. I don’t have much experience with rendering.
I know that lighting is significant, so I welcome any good tips on lighting in cycles. What about polygon count? How significantly does polygon count affect the render times? What about textures with alpha?
I’d like to know enough that I know specifically what to keep my eyes on as I put together my projects. For instance, I may make my sets lower poly and design my characters to avoid subsurf.
I appreciate any help. It is hard to find information on these more fundamental aspects of render times.
you should tell us what kind of animation you are after: making charachters within an environment with full GI advanced materials, phisics sims & whists y bells takes obviously more than motion graphics. once we know your field we can give our suggestions to optimize.
One I can tell so far: poly count doesn’t affect much render time, unless your go crazy with billions of triangles.
Number of samples, complexity of shaders, type of shader (for example volume shaders take very long, refraction can take long), amount of stuff only affected by indirect light (because this requires lots of samples). There are gazillions of other things but i think these are somr of the most important.
You can keyframe the values in your modifiers, something that could be useful for stuff that is farther away from the camera. No reason for the character way over there to have the same level of subsurf as the one that is in focus. And this can be done with materials as well if you are a bit savvy about it.
One thing that speeds up a lot when you are using complex shaders is to change them into a simpler version for secondary bounces or diffuse rays… Baking also speed up things.
I am making a 3D Minecraft animation in Cycles. I am still writing scenes and creating a look. I am more of a modeler, rigger, and materials person than anything else. So, I apologize if I am too ignorant in this area.
How does geometry influence render time? Where does BVH come in? What about normal maps? I can look at each one of these individually, but what about them compared to each other? What I’m looking for is something like a Youtube channel or blog that talks about render optimizations during scene creation, especially within context of a full scene. I want to intelligently decide which compromises to make.
Thank you for all the tips. I am very knowledgeable of shaders. My main shader is a fast shader, but I was just thinking about how to implement more complex things like SSS or translucency. Turning them off with a toggle or swapping them when seen by non-camera rays are great suggestions that I can easily implement!
Turning off subsurf at a distance is great, too.
I’m not sure about the cost of baking shading. I would not be able to instance geometry or tile textures, right?
Indirect light could be a problem. Large sets are a problem (poly count?), so I am trying to keep my sets small. In Minecraft, caves are an easy way to do that. My current plans light everything with torches or lava. But, I haven’t done enough full-set tests to see how difficult the lighting is. If I can’t make it work, I may need to think of completely different sets.
Or just render on a render farm and never worry about boring and time consuming things like optimzation again.
Check the renderbot thread for really cheap, really fasr rendering. It is in the Latest News fo
rum. I cant paste the thread atm because somehow my tablet doesnt like pasting in this forum for some reason
Hey, it’s AustinC again. First off guy never attempt to animate over 720p. And, by all means bake what you can in the way of lighting that is not changing. And, if baking a object of any size jump right to 4096 and you won’t be sorry. I just noticed several walls in that little project that were baked at 2048 and need to be redone. And, you already know that anything with a glossy component can’t be baked.
I noticed the other day that clamping adds render time but in that case what can you do. The main thing is unless you have a machine like the Blender Guru animation comes with many trade offs. So before locking your story line in render a few frames to see if you can live with the per frame time. Some people say five minutes is acceptable others ten. But, of course yours could vary from four to twelve minutes as the story plays out.
That along with the other post should get you up and running. A German friend said animation at home is about cutting your losses. I’ve had to accept with my equipment it is never going to look like I might wish. So you learn to bite your tongue and keep plowing ahead. Best of luck with your project.
So, these seem to impact render time the most (in order).
Lighting setup (lamps vs mesh emitters, how the scene is lit, and settings on all emitters)
Render setup (light bounces, caustics, etc)
Shader complexity (some shaders are noisier than others)
Poly count
But, I still question the poly count. I had a test render with just a cube. I put a generated texture on it, and it resulted in high render times. After I baked it to a normal map, the image rendered very quickly. If I attempted to replicate the generated texture using a displacement map with extra geometry, the render time increased significantly. But, mostly all the render time tweaks deal with the first 3, but I understand very little about 4.
Actually, my work in Blender has focused on asset creation, especially rigging. I have very little experience with baking (practically just normal maps). I had never heard of baking lighting for animation before this thread. So, everything you said is useful to me.
It seems like I need to become a master baker.
I am trying to get a 2-4 minute render time per frame. My PC died, and I haven’t built another yet. So, currently I am just working on pre-production work (like writing, determining my look, creating shader groups, creating assets, etc). I posted this thread because I am just about to make some large design decisions to move forward.
I thank you all for your tips. I have been looking at various resources online about some of these things, and I will need to make some big changes. I need to construct my sets in a completely different way than I intended to so that baking is possible. I may also want to completely redesign my sets to favor “safer” lighting until I get an idea of what my final render times are. I am very glad that polys aren’t a severe issue; I won’t have to worry about them while modeling my main characters. That is the silver lining.
You don´t really lose time. You simply activate more nodes. I usually render with 10 or 20 c4.8xlarge units. It gets a bit more expensive but you save time which is money for optimizing.
That is a good point. I was thinking more about avoiding mistakes rather than devoting time toward specifically optimizing a scene. The value of time is different for everyone, and each person should consider it carefully for their specific case.
I am in a rather unique case. My day job isn’t art. My budget PC build died, and I need to finally build a good workstation (mainly for art rather than for work). I am on a tight budget. And, any pay that I have gotten recently for art work has been negligible ($1-$2/hr). Since I need a workstation to make animations anyway, the most efficient solution is to invest a few more hundred dollars in my workstation and to ensure than my projects are small enough to be rendered on it. Waiting for a render isn’t a big deal if the wait isn’t a few weeks.
A concept/prototype scene that I rendered recently used large procedural textures. I baked the procedural textures into diffuse color textures and bump maps. Rendering that scene without baking would have taken an insane amount of time/money. The ~30-60min spent UV unwrapping and baking textures once per scene would easily pay for itself compared to how much rendering time it saved.
On the other hand, baking the lighting seems to be impractical. The texture resolution required to make it look good is too high for me. The render took 1GB of RAM for a small scene. If I won’t bake the lighting, I can reduce the RAM by about 45%, allowing for much larger scenes. Also, the scene used physically based shaders. Not baking the lighting allows the scene to look far better than using diffuse shaders exclusively. Regarding light baking, I don’t consider the decrease in scene size and lighting quality to justify the saving in render times. But, I consider color and bump map baking to be essential. So, there is a balancing in trade-offs.
It is hard to find information on these more fundamental aspects of render times.