Transport Tycoon Forums

XeryusTC wrote:

Killer 11 wrote: You all know that every decent player that plays and builds rail networks always has atleast 200 trains runing and that's trains only then add buildings, factories , trees ,all infrastructure then add all other vehicles then add up multiplayer wich multiplies vehicle ammounts by good 3 or 4 times then you all know that huge train stations handling like 20 trains at once and that's trains, only aren't that rare, actualy theyre damn common.
This will be a huge stress for your gfx card even on a professional and vell optimized engine. I know you all will say we can use LOD system but it takes some time for LOD system to calculate a lot of objects and if they also move rapidly and enter and exit theyre lod level you're effectively killing performance.

This makes me wonder if you ever heard of culling. Culling algoritms make sure that things that are not on the screen/behind other objects are not rendered. This really lightens the load on the graphics card, in stead of 200 trains and thousands of buildings it would only have to render a fraction of that.

Wouldn't this make the load on cpu bigger and as you know cpu in openttd is quite busy moving all those objects already.
And in multiplayer there's lots of places where you actualy see hundreds of vehicles(had one crazy dude buying 150 buses and using them to transfer passenger to hidge trainstation with 20 trains in it and only a few buses were out of the screen.
Also about the rts game thing seriously there isn't many rts game that have such intense movement in rts you don't have so many objects as in ttd all at once in one screen. Warhamer 40k being a game that sometimes gets stuffed with units tends to lag on geforce7600 and we're expecting more stuff and at higher quality in OTTD.
Also premodeled lod levels make it harder for artist as they have to model, uv map and skin their model multiple times. It would help to have a LOD generation plugin with a decent algorythm for blender so we can automate it but then it makes the game size bigger.
And in the end the game won't look too impressive, to make it look impresive engine needs to be powerful and coding a powerful 3d engine may take years.

Hehe, well the CPU problems arrives when people do stuff that the game wasn't orginally coded for. There used to be hardcoded limits after all, and maps rarely got bigger than 256x256, and the old pathfinder didn't really look ahead.

If you do it right, the CPU load shouldn't increase, only the GPU* load. And I suspect that most computers with a GPU is hardly used when running openTTD, since I suspect the CPU do most of the work.

*(Almost wrote PPU, I really need to stop programming for old consoles that much...)

I don't think it's time for OTTD to move to 3d if ever. It's going to be at least 5-10 years before it may be a viable idea. But for the moment i think 2d OTTD plays just fine. The 32bit version will look A LOT better than what can be done in 3d.

Killer11: The gain of speed on the GPU is in all the cases I know of higher then the loss of speed on the CPU with culling.
And it is probably not a good thing to use as high-poly models as in the blender thread, those models have way too much polygons to be shown all at the same time.
There a plenty of good LOD algoritms that do their job very well without slowing things down too much. You just need to find the right balance between caching (memory usage is high) and calculating every tick/frame (CPU usage is high), if you get that you have maximum performance gain without asking too much of one part of the computer.