Here's a quick video of some of the tests I have made to try using planes as background scenery. It does work better than objects, but not quite to the degree you would imagine. Judge for yourself.

It works ok, but not sure how much you would benefit at the moment if you had to get the trees either turning to face you or by adding another plane to give them some depth. Still, as expected this would be faster for use on distant stuff you will never reach.

I used lots of tree planes in Rescue the Princess and although it was a lot faster than full 3D trees, it was not as high performing as it could have been. Dynamic batching would have helped reduce draw calls a lot and really helped speed. I also used a billboard script on mine so they always faced the camera. An alternative would be to do 2 planes in an X pattern.

I think you'd have a better visual result with a better texture. I could send you one if you need it for a project.

I have noticed that FPSCR currently loses a lot of FPS whenever you use partly transparent textures.



-Wolf

what fps do you get if you do actually use trees? i get around 23~25fps with 4000(ish) animated trees in a hill shaped similar to yours with everything on medium and shadows and stuff set to 0 (medium seems to be the highest fps for me in the current build on any map though) if i turn shadows and water etc up to around 10 it instantly drops to about 18fps and if i turn them back up to around 100 it drops to around 10 fps... interestingly enough it's still around 15~20fps with only 2500 trees (draw call increases for the extra 1500 trees but fps doesnt drop much at all)

@ tomjscott. Yes performance is nowhere near as good as I expected from these tests. Let's hope 1.9 is a big improvement. If even our attempts at reducing scene poly count are not having much effect than we are pretty stuck for optimizing stuff in any way

Nice to see I'm not the only one struggling with finding a way to duplicate a very populous forest. From what I've seen from Lee's blog I think we'll see gains, though probably on the order of 20-40 FPS (which is still great, at this point). I'm hoping to see similar tests from him during his next performance tuning cycle.

I read Lee's blog almost religiously, lol. I have seen several posts about his latest performance work which sound promising. Apparently the Escape Demo now runs at 90 compared to 32 fps, so that sounds a decent boost. But the Escape is still pretty small and fairly low density as far as entities are concerned to be comparable to many of our own levels. Chances are on a large complex map we will still have the same problems, just a little bit further down the line than we do at the moment. hopefully though, as shadows are such a resource hog atm, the new speed boosts on top of having a far faster shadow system may help a lot!

yh i did try to upload some pics with my post but while it said they uploaded they didnt actually appear in the post after i posted so not sure where they went
yep it's a geforce 460 and the tree is Tree 03 (animated) from the foliage pack
like i say, medium seems fastest for me, highest is really slow (like 7-9fps) and low is around 3-5fps slower than medium

Yea Tree-3 is great for me also i cant do this with the other trees but with over 1000+ trees i get 60fps on high but turn on the shadows
and it drops the problem for me is the editor its self became to slow lagging if it wasnt for that i could have filled the map

Just done a test and got to about 1300 of em but was down to about 16 fps by then, regardless of any setting I tried. I'm actually not keen on this tree as you can see some glaring seams on it, so don't much use it normally, it also blends in with the lush terrain too much. Not much point trying to go any further than 1300 really, 4000 would be impossible on my machine. I certainly got a better performance from planes myself, but obviously being able to plonk an entire 3D forest you can roam around would be better. Just not viable for my machine as yet.
I find it odd performance is radically lower with my 660 card than a 460. I still half suspect the cpu as well, as all of yours are significantly faster than my Q6600.

is your cpu under full load when using reloaded i looked at mines and it only uses 8% when running the escape.

well i picked tree 3 at random last time so i thought id retry with some others, with 3 different but still animated trees combined (about 3k without shadows) i don't notice much difference in fps (actually seems higher)



but yes the editor is definitely a problem once you get a lot of objects on the map, i can get as low as 5fps with 5k entities... makes it very hard to do anything as everything is stuttering like crazy.

it could be time to get myself a geforce gtx460 lol its so strange why a lower card can match or beat a better card :/ i had this before where almighty hood was getting the same fps as me with a lower spec ahh well.

Not checked the cpu load as I only have one monitor and can't swap to see it, so no idea how much it's being pushed. Does seem odd that a two generation older card of the same range out performs the 660. A 480, I could understand, of course, as they always keep up for a good few years with newer mid range cards. But those are the quirks we have to live with at the moment

Anyone know as its not clear to me how Reloaded now in current versions, how in V109 when its released or in any future versions perhaps (silly question I know). Handles or will handle these issues in outdoor environments?

e.g.

Terrains and Terrain/World object details :

Seems to me that given the world size of a Reloaded level (even if smaller play areas are used) is relatively speaking quite a large area....terrains and outdoor environments are in general often in game requirements needing to be of quite substantial detail unless an environment is largely flat, empty or surface details and the world contents and detail is low. Rarely the case in many instances.

Given that we ever had the ability to and tools to shape terrain allowing a greater control of its features or even as now in editing the more detail the more ploys clearly how is this handled now or going to be handled to improve performance or is it a question of simply either reducing terrain detail and form/features to maintain a level of ploys the engine can handle at whatever the number is and that's an end to it. So if a game level runs slow then the only option is to remove or use less terrain detail/ploys than we are able to incorporate now.

On the topic at hand similarly what method are we using currently to manage detail added by placing entities either dynamic or static?. I must admit I don't know where we are with this at the moment or where we are going? Do we have any kind of instancing/cloning/load/unload or any other method at all of saving on large numbers of entities drain on an engines performance by any means? or is it simply a matter of each and every individual entity placed is calculated as a single individual entity and each instance added increases drain and lowers performance?

As an example test I am currently using my Sci-Fi level. At the moment it is relatively low in terrain detail and with a very few entities placed so is rather barren. I do not have any fps/performance to spare to do much more with it, particularly if I wanted to use the highest quality settings so use all the lowest which wont give an acceptable game quality and even using lowest settings still will have nothing like enough performance returned to continue developing the level fully as its should be.

Such a level could conceivably for best results as one would aspire to, contain a great deal of terrain surface detail and surface features requiring a large number of polygons and or both high levels of terrain detail plus large numbers of entities.

For example Buildings and Rocks and so on. I could certainly need to find a way to simulate tens or hundreds of thousands or millions of rocks, stones and pebbles and so on and some of these obviously would benefit from more detail and polys than others. Yes I know we can achieve quite a lot if the potential exists to add much simulated detail via use of textures and so on to help but still that is limited in the results that can be achieved. Textures on a flat terrain surface for example can only do so much to help.

Currently I have been looking at making and adding my own buildings potential - needless to say with limited detail as far as possible so as to keep polys as low as possible. I am awaiting the Con Kit to see if and how such may be of greater benefit to this end if it will even be able to create a similar shaped building with better performance returns, but how long one would have to wait to get there is anyone's guess if ever. It will need to be quite capable all round I imagine.

I need to also look at adding a lot more detail/polys to the terrain within the limitations of its capability and or and ideally both adding a large number of entities, Rocks and stones and so on. These I will have to make I guess so I can have the ones I want with as low a poly count I can design and get away with.

Not a LOD expert and don't know if that works or would make any difference anyway at the end of the day. Doubt I would have the time to get into that in any case either extensively.

Whatever the point being is that though experimenting and testing, I know not why as I do know that at the end of the day at least given current circumstances I am largely wasting my time as performance and engine efficiency in my case at least will not allow this to happen for me to build a level with the level of detail and content as required. Not with the engine out of the box at least and that usually from experience means success is doubtful no matter what a user does to optimise and still achieve the desired end result.

Waiting for V109 I guess to see where we are and what we have to date.

I don't even know if this is possible, but a possible performance increase would be if the engine could at certain incremental distances (user-setable) render all the 3D objects onto a 2D plain (Or possible an arc to avoid corners) and then cease to draw everything else. If such a method were possible it would allow the whole terrain to be populated with hopefully a higher running FPS.

I suspect plains are used a lot in games, an example being Max Payne 3, where the backgrounds which are seen but not accessible must be.

Some of the latest games do and always have used simple plain images to create more depth of detail at various ranges using staged, layered distances and where they can be of benefit. I used them in Classic and other engines too where they can help.



Where trees and alike are concerned if an engine will support any size plain then one can make them images of say a long area of trees (or a long fence) so can use quite a large looking forest image in the horizontal to cover quite a large area in one go as well as single instances of individual trees. With well made textures they can do a good job at some distance from player especially when performance gains are vital. A trade off between performance and quality somewhat perhaps but when fps must be gained they can help.



If you use a mix of models and plains and additionally use the plains at varying distances near the extremities of player proximity to boundaries and so on especially in outdoor environments of course then additionally adding similar visual imaging to the Skybox image to blend in these plains and continue to reflect the plain images into the horizon that can help further where that fits in sensibly where scale is not an issue.



If you can in your game also benefit from a little fog thrown into the mix which is not always appropriate perhaps, but if it is then that can also help smooth the plains harshness if that's a problem as long as the plains can accept fogging of course without misbehaving.



Currently Personally I have not seen many examples of user for instance adding a continuation of the real terrain and terrain details and applying an extension of the real world objects onto the skybox as an image to give an appearance visually to add an additional feeling of greater world size/distance and depth/scale of world whilst adding to the apparent detail in the distance.



Of course those things do little to help in areas in close proximity to the player to save on performance if you don't want plains near the player to be easily identified as such so models have to be used. Plains however certainly may be helpful at distance.



Ideally plains should also cast a shadow where they might be seen which may be problematical and or also still add to drain on performance if there are a lot of them.



In theory they should perhaps take up less performance resources but I am not sure they have made much difference in engines of the past where the problem has always been that even the use of plains have often been to much for an indie engine when there becomes a need to add a lot of them. In that case the engine just needs to be much more powerful to even get much benefit from them. Saving a few fps when you have 30fps is not much help being the point. When you have 300fps every one counts to keep you away from getting to the 30fps and below.



Not sure what happened to the quads we saw earlier but they were OK one version and not the next. I guess we have moved on from all that now to what comes along in V109 so best wait and see from there onwards where it all goes.

How does Skyrim get away with so many populated trees in the scene? I noticed from a distance they do billboard.. but there must be something more going on. Do they drop their shader effects perhaps? I think there is something we could learn from that game.

Skyrim must use plains in the background, or quads, based on objects when closer. But it uses other tricks to keep up speed, as all engines do. If you look carefully you will notice the real time sun actually stops and starts continuously. It waits say 10 seconds, then moves a bit, rinse and repeat. Generally you don't notice this at all, but on a general FPS saving note I imagine that it helps quite a lot.
For me this test has proved that the number of entities in a scene has more impact generally than the amount of polys on a model by the looks (within reason). As said above and mentioned in my video, a bigger plain with more trees would also help keep speed up, but of course that can have it own difficulties. Flat floor great, but not good for hilly scenes as much. Plains do cast shadows in Reloaded but due to their one sided nature, only from one side, and yes they slow things the same as any other objects casting shadows.
We as users perhaps can't do a lot to improve matters as much as we would like, and so can only rely on new updates to give decent speed boosts. However, as I am sure all agree, 100 plains, should run a lot faster than 100 complex trees. They run a tad faster, but not as good an improvement as I hoped.

Has anyone yet tried a similar scene with trees that don't use transparency (dead trees) and see how the performance varies from trees with transparent planes?



We gamemakers are saying the wierdest things, don't we?