Hi Daniel,
Based on my recent benchmark I'd say there shouldn't be much difference in overall performance between fixed and programmable pipeline, at least not on Windows OS. The benchmark uses standard lighting and phong/blinn materials, with the lighting done in fragment shader. The shaders are auto generated with osg shader composer (no branching at all). With the GPU bound scene I've got: CAD Model: Vertices: 9900000 Faces: 49500 Display lists OFF, vertex buffers ON ------------------------------------------------ AMD Radeon HD 6900 Series Fixed Function DrawThreadPerContext: 83.7974 fps, warm up 0.355743 s With Shader - Per fragment lighting / Phong DrawThreadPerContext: 83.2916 fps, warm up 0.321948 s GeForce GTX 970/PCIe/SSE2 Fixed Function DrawThreadPerContext: 58.0693 fps, warm up 1.57383 s With Shader - Per fragment lighting / Phong DrawThreadPerContext: 57.2264 fps, warm up 0.338653 s Robert Milharcic
Hi there I have experienced that generally as soon as I use shaders, my DRAW time starts to increase significantly, and this even if I just reimplement the fixed function pipeline basic functionality. I wonder if some out there make the same experiences. Either it is my poor shader programming (i know about if's and branches that kill performance) or is there really a price for the added flexibility? I'm running a GTX TITAN so usually there is plenty of power... Cheers, Daniel
_______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
