Home› Technical Talk

One mesh modeling or combine mesh and attach it simply

connect2infos
Offline / Send Message
connect2infos
Hi,
my question is:
what's the best way to model for realtime even if that's not really a game but it's a realtime archviz, i must have only one mesh for the object as an exemple a table the desk and the leg must be a cylinder from a box or i can make a box and on top of it a cylinder and then simply attached but without caring of the topology of my mesh? like the exemple
thank's for response
· Share on FacebookShare on Twitter

Replies

  • LaurentiuN
    Options
    Offline / Send Message
    LaurentiuN interpolator
    If you talk about the high poly then you should do it like they are built in the real world, if its the low poly then you can use any technique as longs it looks good.
    · Share on FacebookShare on Twitter
  • phenix3d
    Options
    Offline / Send Message
    phenix3d polycounter lvl 5
    s1dK said:
    If you talk about the high poly then you should do it like they are built in the real world, if its the low poly then you can use any technique as longs it looks good.


    But i'v read that there is many porblem with unwarap and also with the baking of the lightmap if there is many geometry crossing eatch other
    · Share on FacebookShare on Twitter
  • 2bytes
    Options
    Offline / Send Message
    2bytes polycounter lvl 2
    Depends on the game engine,  but modeling the low poly as a single mesh is more efficient with UVs and better for baking maps.  Its more work...but is better.
    · Share on FacebookShare on Twitter
  • Deadly Nightshade
    Options
    Offline / Send Message
    Deadly Nightshade polycounter lvl 10
    I would say it doesn't matter. From my little understanding of how the rendering pipeline in a graphics card works, the amount of sub-meshes doesn't matter. What matters though is points. Everytime you don't merge vertices together, and everytime you have hard edges/extra smoothing groups or different materials on two sides of an edge, you will get extra points - points that the graphics engine has to do calculations on before drawing the picture on screen.

    Take this example:
    You have a plane consisting of two polygons (quads/faces), 4 triangles, 7 edges, 6 vertices. Just a basic plane with a split in the middle in other words...
    If the center edge is smooth, and if all polys have the same material, you will get 6 points.
    If the center edge is hard, and if all polys have the same material, you will get 8 points.
    If the entire mesh is triangulated and have hard edges everywhere, you will get 12 points.
    If the entire mesh is triangulated, and have smooth edges everywhere but different materials on the triangles, you will get 12 points.

    So all in all, what drags up the cost of transform-bound geometry is: number of materials, number of smoothing groups (hard edges) and UV's that are not sewn together.
    · Share on FacebookShare on Twitter
  • connect2infos
    Options
    Offline / Send Message
    connect2infos
    Deadly Nightshade said:
    I would say it doesn't matter. From my little understanding of how the rendering pipeline in a graphics card works, the amount of sub-meshes doesn't matter. What matters though is points. Everytime you don't merge vertices together, and everytime you have hard edges/extra smoothing groups or different materials on two sides of an edge, you will get extra points - points that the graphics engine has to do calculations on before drawing the picture on screen.

    Take this example:
    You have a plane consisting of two polygons (quads/faces), 4 triangles, 7 edges, 6 vertices. Just a basic plane with a split in the middle in other words...
    If the center edge is smooth, and if all polys have the same material, you will get 6 points.
    If the center edge is hard, and if all polys have the same material, you will get 8 points.
    If the entire mesh is triangulated and have hard edges everywhere, you will get 12 points.
    If the entire mesh is triangulated, and have smooth edges everywhere but different materials on the triangles, you will get 12 points.

    So all in all, what drags up the cost of transform-bound geometry is: number of materials, number of smoothing groups (hard edges) and UV's that are not sewn together.
    thank you very much i will try the exemple that you've explained in your comment and i will reposte for the result, in fact i'm searching the best pipline for production in realtime the best way to model so if you have a documentation of the different way to model in 3d it will be grete to share thank's again 
    · Share on FacebookShare on Twitter
  • connect2infos
    Options
    Offline / Send Message
    connect2infos
    i'v tryed the exemple that you explain but i didn't understand how ? can you make a little video please to explaine the theory of the points in your exemple

    · Share on FacebookShare on Twitter
  • Eric Chadwick
    Options
    Offline / Send Message
    Eric Chadwick admin
    This page should help you
    http://wiki.polycount.com/wiki/Polygon_Count

    Also the articles at the bottom here
    http://wiki.polycount.com/wiki/Rendering#Rendering_Articles
    · Share on FacebookShare on Twitter
  • connect2infos
    Options
    Offline / Send Message
    connect2infos
    Eric Chadwick said:
    This page should help you
    http://wiki.polycount.com/wiki/Polygon_Count

    Also the articles at the bottom here
    http://wiki.polycount.com/wiki/Rendering#Rendering_Articles
    thank's Eric Chadwick yest it's very helpful but i want to know more about it i will continue a reaserch about this topic and maby make an other post to share what i find. and if you have any other documentation it will be great to share thank's again 
    · Share on FacebookShare on Twitter
Sign In or Register to comment.