Player mesh with 2000 triangles...
by David Fielder · in Torque Game Engine · 03/04/2004 (2:22 pm) · 5 replies
I replaced the player mesh in the starter.fps with one that has approximately 2000 triangles. In 3rd person mode, the display is quite noticably slower (say down to 10-15 FPS). I was hoping that the engine would be able to handle that many polygons for the main character on my computer.
1) Is there any "optimization" settings that I can make that would be suitable for a game using a higher polygon character?
2) Would this problem be solved by getting a faster video card, or does it have more to do with the CPU (a.k.a. drawing polygons vs. transform and culling algorithms)?
Any opinions on the subject would be highly appreciated,
Dave
1) Is there any "optimization" settings that I can make that would be suitable for a game using a higher polygon character?
2) Would this problem be solved by getting a faster video card, or does it have more to do with the CPU (a.k.a. drawing polygons vs. transform and culling algorithms)?
Any opinions on the subject would be highly appreciated,
Dave
#2
I may also consider implementing an option that goes like this...
if (refresh rate is too slow) then
Use one LOD level down
That would be kinda kool, eh? Has anybody already implemented something like this?
Dave
03/05/2004 (10:53 am)
The PC is an Athlon XP 1800 with a GeForce 2 MX 64. I've been optimizing the mesh and implementing the LOD and it is getting faster.I may also consider implementing an option that goes like this...
if (refresh rate is too slow) then
Use one LOD level down
That would be kinda kool, eh? Has anybody already implemented something like this?
Dave
#3
03/05/2004 (11:11 am)
Don't expect much from a GeForce 2.... 
#4
Second, it's well powerful enough for what we're discussing here, as Torque runs fine on first gen GeForce and Radeons, and a bit less fine on non T&L cards with multi texture capabilities.
What was your framerate before switching chars ?
Are you using LODs with your model ?
ie 2000 triangles is not that much more than the Orc model, and if you're seeing a heavy framerate drop when switching the models, even using LODs, etc. it might be something with your model...
Of course, getting a better video card will bring you a better framerate increase than getting a faster CPUs, as this seems a case of your card not pushing enough vertices/pixels rather than your CPU choking on the scene.
03/05/2004 (11:43 am)
One, a GeForce 2 MX is not a GeForce 2. More like a faster GF1 : remember, MX equals low end consumer hw.Second, it's well powerful enough for what we're discussing here, as Torque runs fine on first gen GeForce and Radeons, and a bit less fine on non T&L cards with multi texture capabilities.
What was your framerate before switching chars ?
Are you using LODs with your model ?
ie 2000 triangles is not that much more than the Orc model, and if you're seeing a heavy framerate drop when switching the models, even using LODs, etc. it might be something with your model...
Of course, getting a better video card will bring you a better framerate increase than getting a faster CPUs, as this seems a case of your card not pushing enough vertices/pixels rather than your CPU choking on the scene.
#5
But... if the comparison between the "stock" models and your model is drastically different then it is indeed your model. Make sure that your col-1 is stupid-simple. This might take some creativity on complex shapes, but the collision mesh is rendered exactly like a normal object. If you simply duplicated your model and scaled it up slightly to make the collision mesh its rendering 4,000 polys, not 2,000 polys.
BIG difference.
03/07/2004 (11:29 am)
Ditto to the GF2MX info. Had one, trashed it. FX5200's are $99.But... if the comparison between the "stock" models and your model is drastically different then it is indeed your model. Make sure that your col-1 is stupid-simple. This might take some creativity on complex shapes, but the collision mesh is rendered exactly like a normal object. If you simply duplicated your model and scaled it up slightly to make the collision mesh its rendering 4,000 polys, not 2,000 polys.
BIG difference.
Torque Owner Stefan Lundmark
Almost always, you can optimize a mesh by 20 % without noticing any difference.