Creating an army in TSE

#1 Well as low as possible. Also going to depend on your hardware and objects. I think the only way to find out would be
to put in placeholder characters and see what happens.

I don't know if there are any specific numbers. Gonna depend on your target client.
A P4 1.6 with a 5200 and 512 is gonna run TSE client differently than a P 4 3.8 Ghz with 2 gig's and SLI 7800's.

www.garagegames.com/mg/forums/result.thread.php?qt=32140

#2 Keeping polys count less then 2k for each char should be okay..
And yeah, it depends on the HW used to run it..
with my P4 3GHz 1gig ram and ATI 9600pro 256ram/128bit I have stable 40-50 fps with about 25 chars on screen at once.

EDIT: Torque Game Engine Documentation

#3 The major problem here is in a typically populated mision (atlas, amazing water, a few interiors, trees, nature, players running around, etc..), TSE, in its current state, really starts to bog down after you start adding alot of mesh objects. Even more so if they are animated. (the vertex skinning code isnt finished yet).

The good news is you can get around all this is by batching the army men together, and adding some type of culling to your new Army class.

Look at some of the shape/foliage replicator code and AIplayer. Also, iirc the foliage rep does a nice little quadtree implimentation with culling, and has an in depth explanation about whats going on by Melv.

good luck

edit: As a side note there has been some headway in porting RTSpack to TSE. That might be something to consider.

#4 I am running into problems right now becuase i am trying to load thousands (10k+) objects into a scene. Most of these are dif objects but i bog down after 50 or so due to the number of objects i think. I hope that an optimization that will be made is how TSE handles a mass amount of objects. If they can improve this then i really think this engine could outperform any other.

#5 Lol, 10k objects WILL bog down any engine.

#6 Well it's not going to outperform all the others in term of speed. But then again you can't get a license for SWG's engine can you?

I for one am counting on Moore's law for the future. Do you have LOD's on your complex objects? (hum can dif have LOD??).

#7 Yes! .dif can have lod..

In our build we have implimented occlusion culling in the scenegraph, and in a densely populated mission it helps tremendously (i'm seeing a 50-100% increase in fps accross the board) . Its still not performing like we would like it to (still someplaces are in the teens), so I've got some more ideas for optimizations. Hopefully when i get em all implimented I'll post a resource.


- for those (like me) who had no idea what moores law was..from wikipedia:

(m

#8 I was reading that Nvidia got 10-20 % performance boost with the last GPU revision for the 512meg 7800 GTX. As an example of how Moore's law is still a truckin on down the road.

#9 TSE is really not optimal right now in terms of batching up geometry across objects. We do have a solution that we hacked in quickly to Marble Blast Ultra and it gave us a 5x increase in speed. We're going to put a more polished solution into TSE, hopefully early next year. Should help a lot. I don't know about 10,000 objects though.