Have anyone experienced performance problems when binding FBO’s in recent NVIDIA drivers?
We have a depth peeling algorithm which switches between multiple render targets. When upgrading from 81.95 drivers to 93.71, the time spent in glBindFramebufferEXT() increased from 5% to 43%, while glClear() increased from 4% to 31%.
I’m not sure whether this is a bug in the driver, if the latest drivers have been optimized for different usage, or if there is something in way we use the API and the extensions that is not correct, and that the new drivers are less forgiving.
glCheckFramebuffersEXT() indicates that the FBO’s are valid.
In one of the examples that come with Open Scene Graph (osgprerendercubemap) there is an identical performance problem.
Thanks in advance,