Thanks to all who try to help!

Are you using LOD? What is the poly count of each tree? Do you know that adjusting the camera focal angle will also affect how far you can see? Are you using fog and setting clip_far to a realistic range? How dense is the mesh on that chunked terrain? What texture sizes are you using (256*256 is the most efficient, 1024*1024 and higher are a significant performance hit).

I don't think grouping your trees into one model is the answer, simply because they will surround the player, and if any one of them is within the view the whole lot will get sent for rendering.

I believe both of you are right. But in the meantime, it seems as though more entities = slower fps. This will prolly change once A6/A7 gets some sort of scene management. Once thats in place, it would probably be a good idea to split your models up nicely instead of going with one large model. At least that's how it should be done, but speed wise shows otherwise as of now.

Yes. William is right. There are some disadvantages when you use one single model with 500 plants (5000 polies) intead of 500 plants with 10 polies each.

When you use fog and clipping then the entire 500 plants will get fogged or will disappear. That can look very strange.

Another solution is: You let the engine do the clipping and the LOD. Damocles did a good exsample of this. You create for every plant / tree a few LOD models. The last LOD stage is empty. LOD 2 and 3 are getting smaller, less saturated and a bit transparent. Then the blending will look smoother. When the model gets to LOD stage 4 it disappears completely and will not be rendered. That allows to display several thousands or even more of plants and trees without problems.

And the best of it: You need no action or script that would make it slow.

Hey fellas...
-You guys are awesome...this fourm is equally awesome...so is 3DGS...
-Germany, Canada, USA, Japan, Sweden, Thailand...
-I wish global Goverments worked together so well!
-But that's Politics so I guess we shouldn't go there...

So, back to the topic at hand...
Here is the situation...
-My terrain is incredible...
Made from a USGS .dem and converted to a .hmp
These are the stats:
1.) 1551 x 1142 = 1,771,242 Verticies... (Thanks to Ver. 6.40 Chunking...)
2.) 1.935 x 2.629 triangle size
3.) Terrain size 2999.9 x 2999.9
4.) 683 x 683 Skin Size (I would prefer a 2048 x 2048 - best detail.)
5.) Covers 30 kilometers x 30 kilometers (Not my largest project.)
-There is a sky dome.
-Cyberbabe is the player...
-No sprites - yet, no Buildings - yet, no snow particles - yet and no AI - yet.
-I have tried many concepts for the trees...
From LOD_0 (140 vert's, 168 faces, 32pix x 32pix skin, 24 bit, 96 res, .bmp)
to LOD_3 (14 vert's, 24 faces, 8pix x 8pix skin, 24 bit, 96 res, .bmp)
-I have tried 1 tree at lod_0 to 10 trees at lod_3 (more trees at distance)
-I have tried the fog but to look natural no frame rate increase...
Too much fog does not look natural...
-My terrain is mountanous so to leave lod_3 empty, when you look at distance
looks bad and when the lods kick-in trees appear and or disappear - bad...
-To get proper scale, my terrain is huge and my "player" is very small...
-Camera distances are therefore adjusted so in 3d person I see the player
properly.
-LOD settings are dependant on the clip far variable...
-"Clip far" is set high... To not loose distant terrain...
-"Var loadfacters" are set very low... to kick in to not see changes...
-If "load factors" were not dependant on "clip far" - this could really help.
-Here are my script variables...
var video_mode = 8;
var video_depth = 32;
var video_screen = 1;
var terrain_chunk = 180;
var d3d_triplebuffer = 1;
var time_smooth = 0.99;
var tex_share = 1;
var pos_resolution = 1;
var enable_polycollision = 0;
var shadow_stencil = on ;
var d3d_lightres = on;
var floor_trace = 0;
-Here are function main flags and variables...
tex_share = on;
video_set(sys_metrics(0),sys_metrics(1),32,1);
Camera.clip_near = 5;
camera.clip_far = 100000;
shadow_range = 2;
sky_clip = 40;
var lodfactor[3] = 1, 2, 10;
sky_color.red = 30; //Sky dome is transparent
sky_color.green = 30;
sky_color.blue = 238;
camera.fog = 1;
camera.fog_start = 0.01* clip_range;
camera.fog_end = 2.5* clip_range;
d3d_fogcolor1.red = 132;
d3d_fogcolor1.green = 132;
d3d_fogcolor1.blue = 234;
fog_color = 1;

-I will be using a 3 image splash intro...
-I resourse (.wrs) to pack files and shorten game load time and file download time.

-I am working with a new Sony Vio w/ a gig of ram and 2 gig processer, water cooled - dual processer, so the build compiler kicks in to the 2 cpu setting...

-Maybe I am asking to much of the software and of users computers!!!!
-If you look at my "ski game" (link in my signature below) you will see
that I offer internet downloaders 3 quality versions dependant on user
computer configuration.
That game does not use a USGS .dem
I made the terrain from my memory and site analysis.
I had not yet figured out the .dem -to- .hmp conversion process.

Thanks to all trying to help...
If we figure a way to make this work well...
We stand to advance global simulation use...
Thanks again...
Paul

First, check the bottleneck of your level. When pressing [F11] you're getting the CPU time for different tasks, like rendering and script execution.

If model rendering time is the bottleneck, try to simplify the LOD2 and LOD3 stages of the trees. Group 4 to 8 trees together in one model. Make sure that the tree models don't have animation.

If script execution is the bottleneck, check whether you can speed up your scripts. Of course, you should definitely not assign any action to your trees!

Also, assign the my.dynamic = off instruction to an entity that doesn't move or otherwise have any actions (such as wall or nonanimated decorative object). Things imported in WED (both entities and terrains) seem to be dynamic by default. For this, all you need to do is add and assign this one-line action (more lines if you want to assign a pointer to it):

Code:

---

action make_not_dynamic
{
	my.dynamic = off;
}

---

This has little effect with low-poly models and terrains.

Edit: left out some details.

I'm not sure this will do any good, but you could possible measure the distance from player to tree, and then, depending on how far morph it with a lower poly model, or even a smaller skin. The greater the distance the less polygons your models will have. It will still look good because you're too far to see a difference.

You could also change alpha with distance as opposed to fog, like in half-life 2.

That's otherwise the same system I use for LOD - using vec_dist to find the distance. Instead, the invisibile flag is used for objects in the same position as it's much faster. ent_morph uses the hard drive and is very slow. With this sort of system, you get far more flexibility with LOD than what the engine has built-in, far more.