my 5 cent to an old dicussion -> about the vertex amount in games or any realtime application.
meanless or not? of course the vertex count, means each vertex must be produced by its position and in option -> color, normal, binormal, textureUV ,on one hand on the cpu or since geometrie shader got introduced on the gpu (way faster). so the conclusion is, this is taking time. now the question is how long does a vertex take to get build by its data on the cpu. (ok to make an more serious point id to run a demo program to messure this, but im to bored too.) as you can imagin this would coast a highly tiny amount of time. i guess 1 million is builded in about 1 sec on regular computers. so if we going to flood the entire screen in eg. resolution 1920X1080 (hd) it would take about 2 seconds to build all those vertieces to flood the entire screen! but once this buffer is builded it can be used for everything. (drh vertex method model) ok in regular games and in the near future you have to handle lot of vertex buffers for each modell. but either here you can use a frustum culling so the graphics card have only to handle the vertieces which are in screen.
so now the point is, is it cheaper to reduce the amount of verts or not.
since vertdisplacer like mudbox, zbrush and so on are introduced. the handling of a huge amount of verts got easy to done. so it was possible to sculp details into the model without moving a punch of selected verts around. the conclusion - the vert count got higher, before the peoples used only color textures to bring details in. but by introducing of vertexdisplacing programs it also got more into the direction about using normalmaps (which are pixel exact) to make them models more alife. but the normal maps have one big error, they costs alot of space in the video ram too. and the have to been baked before, which coasts alot of time and trouble solving. so the logic conclusion is, dont shrink the amount of vert extra for solving any performance issues, its way cheaper to invest onetime a bit of time in a render optimization than to start to retopo, making the uvs, baking way.
my personal mind is. the vertex count up to a filled screen is meanless to reduce. but a piece of sand would be senseless to build with myriads of verts. depends how far away you observe the sand.
some points of course speaking against.
- the data size
- and religious ideas
Dienstag, 24. Juli 2012
Dienstag, 17. Juli 2012
Samstag, 14. Juli 2012
i did some griptape designs. those fits on some globe cruisers, such as: globe eyestorm, globe kaguya and all the decks with the similar shape.
of course the collection become bigger soon. if you want one, feel free to get in touch to me! firstname.lastname@example.org