Hi, Everyone!
I’ve attached an image to explain what I mean. The first one image has only one triangle but bigger transparent area and the second one image has more polygons but smaller transparent area. So which would be more optimized for unreal and which aproach would be better? Thanks!
IMO, in this case option A should be more performant (assuming this is non-Nanite).
Both layered masked materials and higher tri counts do affect performance, so it’s really a matter of managing both. In your case, option B greatly increases the tri count without saving much masked space, so I would consider it the worse option.
Generally though, both of these methods affect different parts of the rendering pipeline, so the only sure fire way to know is to test it in-level.
Would love to hear others opinions on this though- I feel like there is little actual information on how the two methods stack up.
Hey @Koroldev!
I think I remember seeing something about this a while back- The one on the left is more performant, however the one on the right could still be better to use with some adjustments for things like moving leaves, etc… But on a basic level as a flat plane and no collision issues, left would be better afaik. Try putting a bunch of each in a scene and using Unreal Insights! 
I agree, there really isn’t much to go on other than standard practices, it’s really dependent on your materials, etc. Your mileage may vary.
Depending on the rendering method used - showing as little transparency as possible could be much more performant.
Large transparent areas could cause massive overdraw. At least that’s what I heard a few times. It’s still not very clear for me, it could differ depending on if you use Nanite or not ?
I am not going to usu Nanite. And this will be a tree with a lot of such leaves.
So I am going to use picture 2 as LOD0 and picture 1 as LOD1.
@sarahlenker @Mind-Brain @Sebastian
Thanks you all for the advises. I’ve made some tests with a lot of such leaves on a scene.
Case 1 with one triangel gives about 30 fps.
So in my current situation it does make sense to use less polygons, Thanks!
That’s a nice test. You said you won’t be using nanite, but can you just turn on nanite temporary for both cases to see if there’s a difference ?
This! In fact, it’s likely that enabling nanite will rocket your FPS up, it’s so useful for things like this with large amounts of models. I’m interested to see that, as well!
Sorry, I’ve already deleted the example. My aproach is to using as less polygons as possible so there was not needed to use nanite also.
Yeah, to follow up, Nanite really doesn’t do well with masked geometry. In order to run a proper comparison you would need a third version of the mesh that is fully modeled.
However, since Nanite is an entirely different system, even though that’s probably the closest you could get to a 1:1 comparison, it still wouldn’t be completely accurate, since there’s a lot more that goes into Nanite and its accompanying features
