Opencv video caputre using gstream rtspsrc slower than just using opencv

Hi, I have been beginner in gstreamer and trying to run a custom object detection model (tiny-yolo) using tensorrt engine on jetson nano. For the input I have a rtsp camera stream.
I want to use Gstreamer for hardware accelerated Gstreamer. I am using following gstreamer pipeline

“rtspsrc location=$rtsplink latency=10 ! rtph265depay ! h265parse ! omxh265dec ! videoconvert ! videorate ! video/x-raw, format=BGR , framerate=60/1 ! appsink”

with opencv
cap = cv2.VideoCapture(

while True:
ret,frame =

When i test the time for capturing the frame, without using gstreamer the time is around 20ms but with gstreamer it is 80ms. Is something wrong with the pipeline? Or am I missing something?

Thank you in advance