Feature Detection using OpenCV 3.3 and Jetpack 3.3 takes longer than OpenCV 2.4 and Jetpack 3.1

I have recorded the compute time to detect ORB features on an image using two TX2’s, one with Jetpack 3.1 and OpenCV 2.4.13 and the other with Jetpack 3.3 and OpenCV 3.3.1. Here are the results:

OpenCV 2.4, Jetpack 3.1 Feature Detection Time = 35.96836936936937 ms
OpenCV 3.3, Jetpack 3.3 Feature Detection Time = 38.265005005004994 ms

I have verified that this is not an issue with OpenCV as on my laptop, OpenCV 3.3 runs faster and takes less time to detect features. The code to initialize ORB features are as follows:

OpenCV 2.4
cv::OrbFeatureDetector detector(1024, 1.2f,16,16,0,2,cv::ORB::FAST_SCORE,32);

OpenCV 3.3
cv::Ptrcv::ORB detector = cv::ORB::create(1024, 1.2f,16,16,0,2,cv::ORB::FAST_SCORE,32);

On both the TX2’s, OpenCV was installed using Jetpack. Any suggestions on what may be the issue here?