# “Updating Objects Flags” — Best ways to get around long rendering times?

I'm currently trying to render an animation that taking 15 minutes to render one frame. Whilst watching the render, I can see "Updating Object Flags" stage seems to be the main culprit. The scene is not that complex but has two particle systems.

Q: Could anyone give me some tips on the best ways of getting around this and reducing render times?

• You can try steps described at blender-guru blenderguru.com/articles/13-ways-to-reduce-render-times – Photon Sep 17 '15 at 12:58
• Thanks Photon. I've found this link already but the tips only make a very small difference :( – alphaman_9 Nov 19 '15 at 6:32
• Are you by chance using volumetrics? This can result in an $O(n^2)$ run of ObjectManager::device_update_flags. – Robert Gützkow Jul 30 '20 at 21:28
• I am not sure about @alphaman_9, but I am using volumetrics, and pretty intensly too – Nate_Sycro27 Jul 30 '20 at 22:08
• @Nate_Sycro27 The only (obvious) way for reducing the runtime of this particular function is having fewer volumetric objects for which the intersections are computed. – Robert Gützkow Jul 30 '20 at 22:33

## 1 Answer

There are a couple of things you can do to try to decrease render time. For example, you could try decreasing the number of volumetrics by using the glare node rather than volumetrics to create glares around lights, by faking clouds with textures or images, by replacing 3D explosions with 2D explosion special effects, by creating hazes with planes that have a transparent cloud texture on them, or by simply eliminating some or all of the volumetrics altogether. It all depends on what you were using the volumetrics for. Of course, it won't look as good as with volumetrics, but that's the price of shorter render times.

Another thing to try is using the denoiser node and reducing the number of samples you use (and the number of light bounces). For best results, check "Denoising Data" under Properties -> View Layer Properties -> Passes -> Data. However, you're still beauty for time.

Whatever you try, there's ultimately no solution for long render times except progressing Computer Science research and render farms (hundreds or thousands of high-end computers working together to render animations). Actually, 15 minutes is pretty good, even for an animation. It takes Pixar about twenty-four hours to render each frame for their animated films. The only way they can render the entire animated movie in a reasonable amount of time is by using huge render farms of thousands of computers working together.

The good news is that there are several online render farms that you can use. The bad news if that they cost money. However, they're still a viable option (better than half a year of rendering for you). You can read more about some of those options here:

Hope that kind of helps, and sorry I can't give you a better answer.

• I am currently using Sheepit distributed render which makes most of my cycles animations possible. I have considered renderfarms, but like you said, they cost money and this animation would be intense. (6000 frames and 24 hrs a frame). I have decided to render the scene without the particles, and keep most of the volumetric planes. I also set the branched path tracing samples to 20 for volumetrics (which can be fixed with intel denoising), and from 120-1000 for subsurface, etc. Producing a total render time of about 1hr per frame (on my pc). Thanks for the help. – Nate_Sycro27 Aug 2 '20 at 22:30
• @Nate_Sycro27 I just saw your picture. You should definitely be able to majorly cut down in render times by using textures instead of volumetrics for the clouds. You could probably find some tutorials online that teach you how to make beautiful clouds without using expensive volumetrics. Another option that might be faster but still have the realism of volumetrics would be to use one large rectangle for the clouds and give the volume density a cloud texture. Again, I'm sure there are cloud tutorials that would use this approach as well. – ElliotThomas Aug 3 '20 at 1:56