Speedy EEVEE Tricks and Settings

EEVEE may be designed with speed in mind, but there are several tricks and settings you can follow to speed up the process even more, reducing your render time. Temptation can come to have the best quality in Cycles, but think about your computer’s CPU, think about all the heat it’d have to calculate through!

Even if a fast rendering time is the main ambition of the Eevee rendering engine, that’s no reason to sit in the complacency, and allow it to take its course, without trying to modify the settings or other tweaks to speed up the rendering times.

Now what’s Eevee? By default Blender ships with three rendering engines: the oldest one is Workbench, which is primarily used as a way of previewing your scene, but not really intended as a rendering engine itself, or exporting to video files.

Cycles is an unbiased rendering engine. What that means is that it tries to realistically calculate how the light moves around and bounces around the scene.

The disadvantage of this is that it takes a long time, and as you can see here it creates a lot of noise in your render, unless you put up the render settings very high.

The other alternative you have is Eevee. Now this system is a biased rendering engine. It renders much quicker. It’s also less accurate though, but if you’re rendering anything longer than a short animation, or if you don’t have the time to wait.

It’s the best option for exporting a video, so it can be uploadable to YouTube or whatever you choose to do with it. Now even if it’s much quicker there’s still some steps you can make to speed up process of rendering, and make it run even faster.

Here’s a little scene i made a few days ago for another video on the channel. Now some of the settings we’ll be mentioning here are situational in nature. That means you have to judge for yourself whether your scene needs these things, or they don’t need these.

One of the first things we can take a look at and one of the easiest ways of speeding up your render times is by changing the number of samples. Broadly speaking the number of samples your render has determines the visual quality of the lighting.

Now on a default render quality of 64 we can see the lighting here like this: it’s very smooth it looks good. The pixelation is a part of the texture, it’s supposed to be like this. It’s a stylistic choice of the video.

But you can also put it down very low. The lowest we can put it is one second. See here it renders almost instantly. By selecting a slot here you can easily switch between different render previews. See in the corner here how much time it took to render each scene, so from three seconds to half a second.

That’s quite a decrease in render times but as you can see the shadow has become jagged as a result. It’s down to you to find a middle ground for what looks good but doesn’t take too much time.

The rule of thumb pretty much can be described: as quicker render times as you can manage without making a noticeable negative effect on the visual quality. If your scene uses lighting, which is very wide in terms of its distribution, then you’ll see that it wants to create this very smoothed out soft shadow on the ground.

Having the render sampling down too low you can see it destroys this smooth shadowing. Even if you put it up to 20 in some cases, it will still create these visible lines where it’s trying to smooth out between the different shades the light is casting.

Now in a still image these lines might not look that bad, but you have to bear in mind that when you render these lines, they are going to jitter around and you’ll be able to see that contrary wise to that if your scene doesn’t make use of any soft shadows.

You can afford to turn the sampling size down very low of course. This can cause an issue in some other places, such as in the screen space reflections and in the ambient occlusion, although i personally don’t use this.

But that’s for you to figure out for your own scene An excellent cinematic tool you can use is the depth of field effect. This creates this kind of blurring which gives the impression of the eye focusing on something in the distance, or in the foreground.

But the greater your depth of field effect is turned up, the longer it’s going to take to render out these scenes. However you can cap how much blur the depth of field effect will render by coming here to the depth of field settings and of changing the max size.

100 is the default. If we compare the render speeds we can see there is a noticeable difference in the two renders. Slot one with a max size of 100 and a render time of two and a half seconds.

Slot 2 with a depth of field size of 20 and a render speed of 2 seconds. Now this is essentially capping how much blur there’ll be so if you have something here in the middle ground, with some depth of field blur, and the middle ground area has as much blur as it’s possible to have, as defined by this setting, then it doesn’t matter how much further away something else might be, as it’s going to blur by the same amount.

That’s potentially an issue that you have to keep in mind. In terms of your lighting quality, one thing you can make use, of something I’ve already made use of here in this scene, is to make use of a light probe.

If we go to add an object and put a light probe, a radiance volume. Scale this up to encompass as much of your scene as you require, depending on what you’re trying to do.

Now what this allows you to do is use the same technology that is used in Cycles, where it tries to accurately calculate the lighting of your scene in terms of how light is interacting with the other objects the lighting bouncing off some objects onto other objects.

It tries to replicate that inside of the Eevee engine. You’ll see as it calculates this bar fills up and it does some crazy stuff here. Even if this manages to pretty much replicate Cycles in a way, it’s still limited that this lighting is baked.

Now baked lighting is something that, it’s like it becomes a part of the texture. It becomes an image on the map. It’s no longer really something that can cast dynamic shadows. You see now how much better the seam looks.

The scene was designed with use of this radiance volume in mind to make it look good. In fact i even intentionally put these shadows going across the building to bring out the definition of the light parts and the dark parts of the map.

However just to show you an example of the limitation of this system. If we just add a cube here. You’ll see it doesn’t matter how many cubes i put in this small area, there’s no darker shadow casts in the centre here, than there is outside in the open area.

That’s because the calculation has decided that this area is this specific brightness, and anything else that goes in there is also going to have that same brightness. But that’s just the limitation of how the indirect lighting bake works.

It could still save you some time especially if it’s a scene in the distance, which doesn’t need dynamic shadows so much. Now i say this saves you time, but how does it save you time?

Well if you wanted to replicate all of this lighting quality, you see the soft shade getting slightly dark as it goes down here.

You’d have to add quite a few more lights to the scene in varied places than it currently has. Like you’d have to have at least two sunlights: one for the areas like this the places out in the open and another one for the places in the shade that’s an indirect sunlight, and then maybe that doesn’t have enough quality, and you have to put area lights to soften it up even more.

Adding many more lights like that to try and replicate this… It’s going to slow down your render times for the extra light calculations it has to perform there. Next we come to what might be one of the most important aspects, and that is to cull certain parts of your scene, which aren’t in use. Now let’s just make up a theoretical situation here.

Let’s imagine we just have someone walking down the path here. Now we have an animation that does this. We could add depth of field and other stuff, but that’s not important for right now. But what i am going to point out is that if this is the entirety of your scene, Blender is still going to try and render out all of this other geometry and all of this other lighting as well.

The best thing here is to delete everything which is not either directly on screen, or which casts a shadow for instance. This tree over here even if it doesn’t appear on screen directly, but it still casts a shadow. It still takes part in the scene’s atmosphere, it’s lighting and shadows and so on. So this should be kept.

The easiest way at least for me to delete and cull everything that isn’t on screen, is to just go to your camera view at the start of the render, zoom out a bit, select everything you can see. It’s all selected, go forward a bit and again select everything holding down shift as you draw a box to add to your selection, rather than remove from it.

Once everything is in your selection, you can try inverting your selection. One thing i should mention is: go up to here to your filters and you want to remove lights and cameras empties, but really you can just keep meshes ticked and that means you’re not going to accidentally select lights or other things that you need.

So once you’ve selected everything on screen that you think the scene might need, go up to the Outliner and invert your selection. The key to do this will be whatever you’ve said it to be, but you can find it yourself.

In the scene it still selected the camera and the volume, I just manually deselected those things. And now just to test before we actually delete these objects, I’m just going to hide them so they disappear.

Now watch the scene again. As you can see it looks pretty much the same. It should look exactly the same, that’s the point. Now that we’ve done this we can safely say that we can delete all of this unnecessary geometry, and it’s not going to impact the scene at all.

The same goes for these light sources over here. These can just be removed. Also these ones over here don’t appear to be needed. You should be careful which you do delete, because as you can see here this isn’t a spotlight, this is actually a sunlight.

You can see if i hide this it’s that which is actually lighting up the scene. Now if we look again we can see that everything looks exactly like it did. Meanwhile everything outside of frame is now deleted.

You can also manually clean up some things. These things were kept because they were behind what was visible. These aren’t really needed, so these could be deleted as well.

You get the idea. Now let’s say it was even smaller scope like this big. You can do the same thing with the geometry itself. Open it all in Edit Mode and the same as i did before, we can select all

the geometry, play to the end of the animation select all the geometry again, invert the selection

and just delete it.

Now you have to be careful here. Sometimes it’s going to delete faces which might want to be kept, depends on how carefully through the different frames of the animation, to make sure you’re selecting everything.

But that’s also something else you can do to speed up your render times. If you want you can also just select everything, and do it like this. But that’s kind of risky in terms of what you might lose out on.

The last thing we’re going to look at, going back to the scene you saw at the start, is volumetrics. Now volumetrics is what you see here: this kind of foggy effect produced by the light shining in through the windows.

Now these can have quite a significant impact on your render speed, so getting the settings to that perfect balance of quality and speed can be important, depending on your scene. Much like sample sizes it depends on what you need in the scene itself.

There’s a few different ways you can go about getting volumetrics. One of the quickest and potentially the least efficient ways of getting volumetrics in your scene is just by going into your world settings and adding volumetric shaders, and all that stuff.

But the better alternative is to create a box, a simple box around the area that exactly needs these

volumetrics. Because if you put it into the world shader it’s going to make volumetrics throughout everything in the entire scene, which you probably don’t need.

We’re just putting it around a box constrained into just this one area, so it’s not going to try and calculate volumetrics outside of the scene. Point is that it constrains the volumetric effect to just

this scene here.

Beyond that’s what i suggest doing. Under the volumetric settings, as was mentioned before, the important ones to note are the tile size and the samples. Now a smaller tile size and a larger sample size are what gives a higher quality render.

And as i said about being the same as like samples, it depends on what your scene needs. For here we can see that we’ve got this nice kind of lightness and darkness, which would be lost or would look bad if the settings were put up higher.

Well there you can see that. But putting it up by too much, it’s going to slow everything down. You can see now how much it’s already lagging. It’s not even moving the viewport around. If it’s lagging like this in the preview, you don’t want to know what it’s like when it renders.

The settings it gives you by default is actually a pretty good middle ground, you can mostly keep it on this. One thing you can do if your scene needs it, is turning down the end distance. This affects how far away from the camera volumetrics are going to be rendered.

Putting this down as much as you can afford to can sometimes help. Although if you’ve already constrained to a box area manually, that detail isn’t really important. Otherwise this has just been a little list of tricks i found over my time using Blender that speeds up Eevee’s rendering cycles even faster than what the default settings can provide you.

Hopefully this goes some way to increasing even your render speeds further than what it had before. Something else i should have mentioned about the culling part… The same goes for lighting as well.

What you can do is select your light source, go down to the visibility tab and animate its rendered little tick mark. So that’s it’s a bit of a slow process, you have to do it manually. But if you’ve got a larger scene with a lot of lights, especially if they cast softer shadows which is going through which lights are necessary at what moments, and animating this little tick box to be unticked, when they are not contributing to what is seen on the camera, you know when they’re out of shots.

Just disabling this will save a lot of rendering time. So this has just been a little list of things that might speed up your rendering times. Some things I’ve picked up over my time of using Blender for the amount of time that I have.