I am trying to add occlusion query support to our engine using OcclusionQueryNV. I have to render around 4000 objects at a given time.
When I run an empty loop running for around 4000 times without rendering anything. I am getting around 1450 fps. As soon as I add only glBeginOcclusionQueryNV and glEndQueryNV to the loop, the fps drops to around 150. As soon as I add a call to glGetOcclusionQueryuivNV, the fps drops to 50 although nothing is being rendered.
I believe just a begin/end query shouldn’t be so slow? Am I getting correct results? or is there a problem somewhere? Anyone else experienced anything similar? or my test case is incorrect?
this has been a problem with nv3x for at least the last 6 months with various drivers, nv4x didnt suffer from the (if it is) same problem ive heard (until now).
sorry i cant be of more help
I’ve done some more tests with an ATI 9700pro on a P4 2.4GHz processor.
An empty loop run 4346 times gives me 790 fps. Adding the begin/end querry calls drops it to 194 fps, better than 6800gts 150 fps. Adding the getOcclusionQuerry call drops the fps to 190, still significantly better than 6800gts 50 fps.
It seems there is apparently a problem with 6800gts drivers. I hope there isn’t a hardware problem with this. Anyone from nVidia here?