rasterizer issue (even in wireframe)

Hello blender game artists,

I work on my Cs projects since lot of time, I updated to 2.69 recently (sometime downgrade to 2.64 for single texture mode), I found a good way to optimize areas, and I planned to begin the “real” area1 soon.
I also planned a video about my lastest updates, but I have a huge problem with the rasterizer, even on wireframe.
I searched many way to resolve this, with the help of this forum, myself, and other, but I come to a dead end, and I don’t want to leave certains thing like my charactere and my camera setup…

My computer is pretty bad and I thought it will be good on a better computer but I tested this week and the issue was there too.

The issue is when my character is on screen or partially on screen the rasterizer increase too much,
and I don’t talk about the first sub area. I don’t have a lot of face here (mainly square polygones),
character 4069, ship 1100, sub-area 2320, door 1050, grass 6280, upper-structure 1848, the last two are no collision.

What I already tried (in wireframe) :

  • delete two lights (no visible gain),
  • set all alpha textures to alpha clip (no visible gain),
  • reduce textures size to 128128 instead of 10241024 (no visible gain),
  • all pulse mode frequency on 1 instead of 0 (no visible gain),
  • delete some logic that use pulse or near (no visible gain),
  • delete all ray sensors and raycast (no visible gain),
  • delete unused modifiers (no visible gain),
  • delete some lists (no visible gain),
  • increase camera lens (gain just a little),
  • “middle poly” character instead of “high poly” (gain just a little),
  • deleted lot of actuators activation and desactivation in my character script (gain just a little on overall)

I am sorry I can’t post a .blend, but have you any idea about what can cause this, do you think it can really be the overall polycount ?
I was thinking about mouse position recovering, message sending, text objects updating, the number of objects, charactere rigging, bad scripting, but the subarea have nothing like that and it increase the rasterizer, so I don’t know…

http://nsa34.casimages.com/img/2014/01/04/mini_140104043308992193.jpg

Your last image seems to show a lot of faces. I guess you try to simulate gras/fur with several alpha faces.

Anyway this seems not to be you main character. Unfortunately I do not recognize what the character is. Your description is pretty detailed and let me assume the problem is:

A) with the character
B) with rendering -> faces, textures, shaders (not mouse, scripting etc. as this would be logic)

could it be you have some subsurf on the mesh?

Hello and thank for your anwser Monster,

My character is no collision parented to a “low-poly” dynamic collision box.
Its “middle-poly” mesh object with 4096 faces, one toon material and two 2048*2048 .tga texture for color and normal map.
Its armature has 75 bones and animations uses 4 or 8 layers of actions simultaneously.

There is one modifier on the character mesh for the armature,
one sbdivision x1 already applied on ground and grass meshes,
one subdivision x2 and one displace not display on ground and grass meshes.

Well, I finally decided to reopened a “not so old version” that display a rasterizer around 75% instead of the 80%.
but there is no change on meshes compared between this two versions and the old one has even more scenes,
so I still think a little about scripts and pulse, like viewport, relplace mesh, or too fast updates with a lack of indentation.

Yes I tried a dense enough grass and for something that good to me I used 10 faces of clip alpha grass for one face of ground,
a grass mesh with half of the previous faces give a little gain and it may could be correct for a “middle-poly” of this object aswell,
so keep the older dense enough grass for a future high-poly and/or ultra-poly mesh of this object can be a possibility.

I started to clean some scripts and review some meshes for a “middle-poly” instead of a “middle-poly+”, as character and grass,
and was able to decrease a little the rasterizer,
30/35% before the introduction.
40/45% during the introduction.
70/75% during the main menu screen,
60/65% after the menu,
and fps are mostly around 20fps during the main menu screen and around 20/25fps after the menu.

Now that I think of it, I also tested some “high-poly” sub-areas some time ago instead of my actual “middle-poly”,
so after the optimization of all my logic I think I finally must integrate this setup menu, with selections of lod, resolution, etc.

Well, I think I’m going to try a little video capture of my lastest version before a new update,
then will try to prevent bad scripting and think about lod layers.

wo 2048*2048 .tga texture for color and normal map.
Its armature has 75 bones and animations uses 4 or 8 layers of actions simultaneously.

There’s one problem. Textures are HUGE. There are a huge number of bones as well.

He already tried changing the texture size to 128*128…

Are you sure occlusion culling is on?

Textures are usually not the big deal … except you hit the memory size of the GPU as reloading will eat processing time.

75 Bones are quite a lot - at the scene graph/animations
simultaneos animations eat processing time - at the scene graph/animations

Python code goes always to logic.

I still guess there is something with semi-transparency (alpha) as this is a render killer if used a lot. It does not even matter if the final appearance is opaque or invisible. As soon as it is alpha it hits render.

For testing purposes, load your character into a very simple test scene and check if the problem still exist.

Hello,

Well, I knew since a lot of time that my character could be a ressource eat and I think since this time that it must be the most intensive object of the game, but I wasn’t expected so much.

I tried without the first sub-area and when the character is on screen it’s around 30 fps,
and when there are just some parts of it visibles it’s around 40 fps, as with internal view.

I will try to reduce its amount of polygone for the “middle-poly” lod layer instead of its actual kind of “mddle-poly+” that I can keep for “high poly”, but I mostly think about increasing the required specs because I am not going to reduce its armarture and animations.

About alpha, I always take care about not activate it on opaque meshes objects
I always use alpha sort for two “halo” meshes objects and two “cloud” meshes objects on the main scene,
for 5 small “plane” meshes objects in my hud scenes and 4 “sky” meshes objects in my background scenes.
And during the display of areas I can variously use it, as never or too much like in the first area where I use alpha clip for grass.

Also after test, my new “middle-poly” grass with half of the polygone compared to my prévious seems finally good enough,
and as already says I must try to reduce others objects where it could be some or many useless polygone, as upper structure.

The occlusion culling has never been disable and has a resolution of 128.

http://nsa33.casimages.com/img/2014/01/07/mini_140107074148636930.jpg