STATIC INFANTRY - ACTIVE INFANTRY

[FC_Schaefer]

I have a question regarding infantry. Like most objects there appears to be two different types. Static. Active. I'm wondering if someone could quantify how much more it "costs" in terms of resources when selecting an active infantry object over one that is static? This is beneficial when designing multi-player missions that offer the Combined Arms component in the (800-1,200) object range.

 

In addition, after hearing about the "skeletal animations" I was wondering if there were plans to include some additional infantry. Currently, I'm using Insurgents for Abkhazian and South Ossetian forces. The Georgian infantry is identical to the Insurgent infantry. The American infantry have no anti-tank capability so I'm using the Georgian/Insurgent RPG object. The static Russian Infantry is selectable in the Mission Editor, however, it does not appear in the actual mission when played. In essence, it disappears.

 

Thanks for your continued efforts with the Combined Arms product. I'm glad to hear that you continue to refine and polish this module and look forward to any news, updates, and future endeavors from the EAGLE DYNAMICS team.

[Grimes]

Static objects have no AI and therefore can't move or shoot. They should be used simply for making a scene "look pretty". So non static objects have AI running and are capable of moving/shooting/animating. I can't give a precise cost in terms of game performance between the two, but non-static AI will always cost more, its just a question of how much.

 

Infantry right now are pretty basic and IMO they need a sizable make-over to get the most out of them.

[Robin_Hood]

Maybe you can try editing a mission the following way :

 

1. Start with a basic mission (it can be empty, or have a few units already)

2. Start the mission and check your frames per seconds

3. Place 500 "vehicule" (active) infantry troops (for example, it can be some other number, and the test should be valid with other types of unit as well) - do not have them move around as this may incur further performance hit

4. Check your frames per seconds with those, see by how much they were reduced

5. Remove all those troops, and place the same number (in this case 500 infantry) in static form (I know, placing 500 static objects will take a while, thanks God for Ctrl+V)

6. Ensure that you have the same number of troops (the units list should enable you to do that)

7. Check the frames per seconds again, and see if they are higher than with active units (it should be, I guess)

 

As I said, things then may differ when units are moving, shooting, etc... (not talking about the purely graphical hit that is sure to happen watching 500+ units fighting).

 

I have not tested this myself, so this may be interesting.

What I have tested though is putting an AI OFF trigger on the units (several hundred), and guess what ? It didn't change anything. So AI OFF is not useful for reducing the strain on the server (it is useful for other things though).

[Pikey]

Are you looking for a cost in memory or CPU? CPU is likely to change per infantry member depending on how many there are and the type of processor, all other things being equal (to be confirmed). I'll try testing to see if perfmon can record the differences. I'll see if there's differences per infantry type also, though it seems you've already looked into at least some of this.

[Pikey]

Also FPS will greatly depend on LOD, distance and the content of the graphics display at the time and can also be affected by CPU use which it shares to an extent.

[Pikey]

Test 1

Method

100 infantry moving in an undetailed area of the map. 2 sets of 50, moving 1 waypoint towards each other. Check CPU via process and resource monitor for average CPU.

Data

CPU varied. In the world view, zoomed in and looking at one soldier the CPU moved between 3-6 according to Resmon. However when switching to F10 Map view the CPU jumped to 20-25% avg CPU for the one process.

These numbers did not change if the units were displayed either in first person or map f10 mode.

FPS was capped at 60 in the first person mode and didn't seem to move.

Conclusion: Map mode affects resource use far more than units on screen and max units when in first person view.

 

I could carry on looking into this but really not sure what worth it would have. These test can be made as easily by the OP as their local hardware can be of more value in precise testing.

[karambiatos]

I have made missions with slightly over 900 ai units on it, and the impact isn't at all noticeable as long as you spread them out nicely, and if you have a commander role, make sure he doesnt move a lot of units at the same time, otherwise the server crashes.

 

another perhaps important factor to keeping client performance up is having a low number of different types of units (vehicles) and by extent different textures, that way the texture needs to be loaded only once, instead of having many different types of textures.

Edited October 11, 2014 by karambiatos

[Robin_Hood]

The F10 map view seems pretty costly indeed. I have made some further tests, and here's what I got :

 

- I used an empty map, with one armored Ural as an Anchor, in a low resolution area (north of Mozdok). I placed the units in the general area of the truck, but not quite in view, as I am only trying to test the ressources taken up by units in the mission, not by their graphical representation.

 

- Having 400 (active) infantry units lowered the FPS slightly (I lost 4 FPS) in normal view (F7 view on the truck) - not as much as I thought, but I think this is very much hardware and situation dependent.

- In F10 view, however, I lost 45 FPS !

 

- The results were identical whether I used 8 groups of 50 units or 40 groups of 10 units

 

- Static infantry units* (400 of them) gave a better result, as I had no FPS drop in normal view, and "only" a 26 FPS drop in F10 view

 

Of course, I'm sure all of this is very much hardware dépendent. Also, units moving or fighting may be another story.

 

* I had to use different infantry, as the standard russian infantry apparently didn't work as a static object - is this a bug on my side ?. So I used paratroopers instead. Shouldn't really matter that much.