scott <sco### [at] scottcom> wrote:
> > The optimal situation is that only
> > those polygons which are visible in the current frame are sent to the
> > graphics card to be rendered.

> Maybe that was optimal 15 years ago, but today the GPU is way faster than 
> the CPU at culling polygons.  Especially when you consider the overhead of 
> having to transfer all the mesh data per-frame to the GPU - that is slow and 
> nobody does it.  The optimal approach nowadays is to have as much mesh and 
> texture data loaded in the GPU as possible, and per-frame the CPU can just 
> tell the GPU which meshes to render, based on some algorithm.

  Many games seem to disagree with you.

  For example HL2 and sequels actually drop meshes and use lower-resolution
meshes for scenery which can only be seen from far away, but if you go
to that scenery, it will load objects and higher-resolution versions of
other objects.
  For example in HL2:Lost Coast you can use objects to climb over barriers
which would normally stop you from getting back to previous parts of the
level. If you do that, those parts are very ascetic compared to what they
were previously (when you played through them normally).

  Other games, such as for example Oblivion, generate terrain details
dynamically based on distance: Closer terrain will be more detailed than
terrain farther away. Other games, such as Call of Juarez, also do this
(and this one does it even better than Oblivion), which allows huge visible
landscapes to be rendered.

  The amount of polygons to render matters even today. That's because
today polygons are much heavier to render compared to 15 years ago.
15 years ago you had bilinearly-filtered perspective-correct texture
mapping, light mapping, and little else. Today you have all kinds of
pixel shaders, bump mapping and whatnot.

  Also some polygons need to be rendered from farther to closer, most
prominently polygons with transparencies, or else the Z buffer won't
work properly. I don't know if any graphics card supports sorting of
polygons (especially since there's no generic algorithm for sorting
polygons, not even triangles, without polygon splitting).

-- 
                                                          - Warp

Post a reply to this message