PopcornFX is more about a lot of particles with cool moves, than a few particle with cool materials. (You could do both if you have the time, and when we’ll plug the material particle nodes (Velocity, Size, Age, etc…))
PopcornFX is about making any behavior you want. You have the “classic” behaviors (Physics, Collisions, Attractors…), but where PopcornFX really makes the difference is with its in-house particle-oriented scripting language, very close to shader syntax: it fells more like having a bunch of cool “particle-oriented tools” to be used the way you want (Basic Shapes, Meshes, Curves, Textures, Spatial Layers…).
As a downside, it also means that the deeper you go, the more technical (as in technical artist) you need to be.
And we put a lot of effort in optimizing the PopcornFX Runtime. So you could not only make butterflies flap and fly the way you want, land the way you want, burn the way you want, but also spawn 20,000 of them with respectable performances (easily quadruple the number if you remove scene collisions/raytrace queries).
That said, in UE, we still have work to optimize the rendering. As much as we already did, UE still does not like very much 400,000 of our sparks flying around. But we are working on it.
And PopcornFX might lack some optimizations for big open worlds. For instance, today, you can only LOD “by hand” in the Effects. Or, we batch so much particles together that it becomes difficult to cull them. But we have already began optimizing for big open worlds use-cases in PopcornFX 1.10.
Also, PopcornFX is a middleware, so (as long as you make them Axis System independent) you can run any PopcornFX Effect in any Game Engine that integrates the PopcornFX Runtime.
And having a standalone PopcornFX Editor, with a big viewport, particle-oritented UI, with a lot of particle debugging and profiling tools, is real plus.
I’ll stop there, but there is a lot of little features here and there that can make the big difference (TwoWay collisions, scene queries, broadcasted events, attribute samplers, future seamless GPU simulation…).