Prototyping for mass strategy simulation- best approach?

First of all: I am definitly new to UE.

I dream of making a simulation for mass combat in a fantasy setting.
I have some questions though in which way I should head.

But first I would like to describe my progress so far as to add some light and life to my questions:

My starting point was to implement a rough simulation of some 1000 Orcs vs 1000 swordmen (with a rather rudimentary graphic) in another programming-language based on the D&D rules (tabletop fantasy setting):

https://youtube.com/watch?v=URf1HccJKhE

I also implemented arrows, riders and was able to set thousands of individuals against each other (though these clips where not life-footage):

https://youtube.com/watch?v=BUga17NO9n4

As scrolling around the “map” became an issue I took advantage of OpenGL 1.3 and changed to 3D (and also included flying units and magics):

https://youtube.com/watch?v=1q8R8s04j8g

Representation of the units was still VERY rough and so I included some finer meshes, some of them created with makehuman:

https://youtube.com/watch?v=vyKwWe9czhE

Matter of fact I can not deny that while I was quite content with the “game”-mechanics (in regard of this early state) I wasn´t really satisfied about the 3D art.
Though I am not a guy who needs an realistic presentation in games (as you might have guessed from my above examples) a bit more decent presentation of the individuals might come handy.
I also hoped, that a game-engine might also simplify some other game mechanics (like path-finding etc.).

So I gave UE a try and made my first prototype of 100 orcs vs 100 swordmen (actually they are rather playing football instead of fighting):

https://youtube.com/watch?v=KaEQUGvmxuU

When I did the same with 1000 orcs and 1000 swordmen it came clear that such a simulation would be “unplayable” cause of “lagging” (at the moment). The following footage is made with matinee and not real-time-streaming:

https://youtube.com/watch?v=KaEQUGvmxuU

I replaced the complex meshes with simply cubes but it soon came clear that (the way I did it) about 1000 characters is the maximum in regard to playability (with a modern middle-class GPU).

I would like to make some more experiments with UE, but I simply don´t know the engine well enough to know in which direction I should head to get far better performance while still keeping thousands of characters. I could accept MUCH less detailed graphics and would also accept that the characters are not animated (like chess-figures), maybe also 2D pictures placed in a 3D map, but would like to hear your opinion about it:

  1. Is the Unreal Engine at all the right tool to head for what I want to do?

  2. Do I get (much!) more performance when I change to NOT animated actors instead of characters?

  3. Do I need to change to a 2D representation of the characters to be able to display thousands of units? If yes: Would widgets be the best way to implement them in UE?

  4. Any other suggestions how I could speed up things a lot?

Thanks in advance for your suggestions!

#screenshotsaturday
#unrealdev