Clutter vs Objects?

[Nicholas Bell 0]

I'm running up against the 4 texture limit per segment on my Schmalfelden map, and was considering using clutter plants as placed objects to represent the different types of farm fields.

The major factor determining the in-game FPS is the number of objects in the player's view. Any experts know if clutter rendering is less system intensive than an equal amount of similar objects? I ask in the hope of avoiding having to perform my own testing and consequently delaying release of this new map.

Thanks.

[dmarkwick 261]

I know that the shader responsible for vegetation rendering likes as few types as possible, to implement some coding trickery probably. And of course model complexity has a bearing, although it seems that alpha complexity does not. Will you be using stock plants? If you're creating your own you could bias the numbers in your favour by having models as simple as you can get away with, and make up the details with alpha, and have aggressive LODs that swap in much more simple models.

If using stock veg, then as long as you keep the types of veg down you should see reasonable performance.

I'm not sure, do clutter objects have geometry? I know if you spawn an object as a SpaceObject (for particle work) then you get the visual but not the other geometry like viewblock, collision etc. This might have a bearing on clutter performance.

[dmitri 0]

Don't know about clutter, but object quantity does eat performance.. Even if these objects aren't in the players view, just the fact that they're on the map. It might also be the proximity of the objects to the player. Having too many close to the player vs spread out across the entire island.

Wonder if it would make a difference creating a large square object consisting of multiple plants? Reduce object count yet give the same appearance.

Don't know if it's relevant, but here is cut and paste from a test I did a while back, objects placed in a 1000m radius around player. Positioned in the middle of the objects, aimed at the ground, with only a few plants visible in the players view. One BIS tree type + one BIS bush. No sat map to remove any effect it might have.

no objects

115fps

3000 objects

99fps

9000 objects

90fps

17000 objects

78fps

60000 objects

45fps

Too bad that last figure is where my own island is currently.

[Nicholas Bell 0]

Thanks, gents. Using objects as clutter is not a good idea. Did one small test and a small area of 120 grass clumps dropped the FPS by 3.

Dmitri, I have 280,000 objects on my map and have FPS between 20-60 depending on where I am looking. I am sure that if you try looking at the ground or sky in your tests you will see the FPS jump. Not saying that the total number of objects doesn't have an impact, I just don't think it is the major factor. The small size of your map and the small cell size may have something to do with your results too.

Bottom line is we both are having the same problem - creating a realistic map without the FPS dropping to single digits. I'm willing, like yourself, to suffer a lower FPS for a realistic map. Last night I played reporter and followed a Soviet assault on a town. I have to admit that despite constant work on this map for the last 4 months to the point of nearing throwing in the towel, the experience was literally jaw-dropping. Even at 20 FPS.

I may do a "mono culture" test where I change all the trees to one type to see how that impacts frame rates.

[dmarkwick 261]

Last night I played reporter and followed a Soviet assault on a town.

Crikey, I thought I was the only guy to do this sort of thing

[dmitri 0]

Bottom line is we both are having the same problem - creating a realistic map without the FPS dropping to single digits. I'm willing, like yourself, to suffer a lower FPS for a realistic map.

True. I can suffer a "little" less FPS if the gameplay experience is interesting enough. I have compromised to some degree. Originally at 5 cell. But I rebuilt it at 6 to gain performance. Optimised object count and have decided to decided to forgo clutter for now. It can be quiet barren in places, especially considering it's supposed to be Vietnam (partly why I've given it the general title "South East Asia"). I'm adding some "different" terrain features that I think will make up for the lack of clutter.

Last night I played reporter and followed a Soviet assault on a town. I have to admit that despite constant work on this map for the last 4 months to the point of nearing throwing in the towel, the experience was literally jaw-dropping. Even at 20 FPS.

Sounds fantastic. Fun watching the game play out on something you've created.

Speaking of actual performance achieved during play. Rahmadi island itself might achieve 100+fps, but add a few engaging AI groups and performance drops to the 30s. No matter how we tweak or steal FPS, it's going to drop regardless of the island.

Your map must be quite advanced after 4 months Hoping to resume work on mine soon after a horribly busy 3 months.

I may do a "mono culture" test where I change all the trees to one type to see how that impacts frame rates.

Tried that myself, recorded no real difference in FPS. Your results may vary.

Last night I played reporter and followed a Soviet assault on a town.

Crikey, I thought I was the only guy to do this sort of thing

I do it all the time too. Funny get caught in the crossfire.