High CPU Occupancy when pulling rtsp network stream by appsink drop=true

I would like to jump frame to pull rtsp network camera stream by cv::VideoCapture using following url.

std::string url = rtspsrc location=rtsp:\** latency=0 ! rtph264depay ! h264parse ! nvv4l2decoder enable-max-performance=1 ! nvvidconv ! video/x-raw,width=1920,height=1080,format=BGRx ! videoconvert ! appsink drop=true sync=false

I found high CPU occupancy by this way, and the reason was “drop=true”. I think all the images from rtsp stream still was pulled all, but they were dropped in buffers of appsink.

Do you have some solutions to reducing CPU occupancy?

Since aapsink is native gstreamer plugin, probably we don’t have much to do for improvement. Would need other user to share experience.

It looks to be similar to this:
appsink pipelines using high CPU (#55) · Issues · GStreamer / gst-python · GitLab

Can frames been dropped in some plugin before appsink, e.g. nvvidconv or nvv4l2decoder and so on ?