3D tilesets and other ideas for LoS (from Starforge)

3D tilesets and other ideas for LoS (from Starforge)

Postby Julius » 14 Jun 2012, 20:48

Found this interesting game concept ( http://www.starforge.com/ ):

http://youtu.be/YxBSYit49c8

http://youtu.be/mJLee5zZ4Ts

And though maybe some of those ideas would fit to LoS as well ;)

Not a big fan of Minecraft personally, but this looks cool.
“You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete” - Buckminster Fuller
User avatar
Julius
Community Moderator
 
Posts: 1942
Joined: 06 Dec 2009, 14:02

Re: 3D tilesets and other ideas for LoS (from Starforge)

Postby TheAncientGoat » 15 Jun 2012, 04:01

The thing is, LoS' performance is already far from stellar; StarForge's tiles are awesome, but we need to get the game buttery smooth sothat combat is fun, for us to implement heavier graphical loads.
User avatar
TheAncientGoat
Community Moderator
 
Posts: 518
Joined: 27 Dec 2009, 19:06

Re: 3D tilesets and other ideas for LoS (from Starforge)

Postby Julius » 15 Jun 2012, 16:05

Hmm... where do you think this slow performance comes from mainly? Sometimes it actually improves the speed if you update the graphics ;)
“You never change things by fighting the existing reality. To change something, build a new model that makes the existing model obsolete” - Buckminster Fuller
User avatar
Julius
Community Moderator
 
Posts: 1942
Joined: 06 Dec 2009, 14:02

Re: 3D tilesets and other ideas for LoS (from Starforge)

Postby TheAncientGoat » 16 Jun 2012, 08:04

Increasing the poly count is not going to make it faster :) Sure, better culling, and mesh-retopology will help a lot, and I'm not sure how much of a performance drain the custom model system is accountable for, or if it's mainly the server/client arch that leads to the performance problems, but at the moment I can't run the game due to a nasty bug, and poor nekotaku is the only person capable of fixing it, and hunting bugs is never fun :(
User avatar
TheAncientGoat
Community Moderator
 
Posts: 518
Joined: 27 Dec 2009, 19:06

Re: 3D tilesets and other ideas for LoS (from Starforge)

Postby DrAltaica » 18 Jun 2012, 04:25

TheAncientGoat {l Wrote}:Increasing the poly count is not going to make it faster :)

You using a software renderer? Tris Per Second hasn't been a bottle neck since the 90's


Just the otherday they where talking about Ogre's permanence problems over at Realxtends-dev group.
  • Tundra 2.3.2.
  • Urho 3D, an open source cross-platform rendering and game engine developed by Lasse Öörni from LudoCraft, http://code.google.com/p/urho3d/ .
  • A custom (crude and partial) .txml rendering loop written directly on top of a graphics API tentatively named 'gfxapi'. Gfxapi is an abstraction of D3D11/OpenGL3/GLES2 for different platforms. See http://clb.demon.fi/gfxapi/ .
The tests were run on a Core i& 860 @ 2.80GHz (8 logical cores) with a
NVIDIA GeForce GTX 275.
The scene contains 428 Tundra entities, 510 textures worth 154MB, 233
meshes worth 60MB, and 672 materials. There's roughly 700 batches
rendered, ~600k triangles when the scene is observed being fully in view,
and averaging 831 tris/batch.
Results:
  • Tundra 2.3.2 (D3D9):
    18msecs/frame (54 FPS), consumes 534MB of main CPU RAM, and 17% of the 8 logical cores.
  • gfxapi (D3D11):
    3.8msecs/frame (260 FPS), consumes 156MB of main CPU RAM, and 6% of the 8 logical cores.
  • Urho3D (OpenGL 2.0):
    2.48msecs/frame (402 FPS), consumes 78MB of main CPU RAM, and 20% of the 8 logical cores.
It should be noted that the tests are not exactly fair and equivalent in
all aspects: Tundra is considerably more complete in terms of features
compared to Urho 3D, and especially to gfxapi, which is simply a
custom-built test app that stuffs the vertex data to a GPU, without even
any kind of scene management.
[snip/]
Ogre rendering, we see here that Ogre is losing out, and by more than a
whopping 10msecs/frame.


TheAncientGoat {l Wrote}:Sure, better culling, and mesh-retopology will help a lot, and I'm not sure how much of a performance drain the custom model system is accountable for, or if it's mainly the server/client arch that leads to the performance problems.
Anyone ever compare the Ogre3d and openGL render?
User avatar
DrAltaica
 
Posts: 10
Joined: 26 Dec 2009, 14:49

Re: 3D tilesets and other ideas for LoS (from Starforge)

Postby TheAncientGoat » 18 Jun 2012, 05:25

Yeah, increasing the polys by adding meshes doesn't make a difference, when I was testing it by adding a couple of thousand trees to the overworld scene (maybe a 5 fps drop, but nothing major), I was more just pointing out that it wasn't going to make it perform better either. Increasing the amount of voxels on screen, however, reduces performance dramatically, not sure if it's the mesh stitching algo or what, nekotaku is the only person who knows the inner workings there. It would be a pity to port to a different engine again, though, because that means that a lot of work the last few months was for naught...

The Ogre renderer was faster when I first tried it, maybe it was an fps cap or something, but LoS' old renderer never went above 40 fps whereas Ogre's goes into the 400's.
User avatar
TheAncientGoat
Community Moderator
 
Posts: 518
Joined: 27 Dec 2009, 19:06

Who is online

Users browsing this forum: No registered users and 1 guest