Npp image encode and decode on Jetson TX1

Hi I am trying to decode and encode a 640 x 480 JPEG image (RGB) on Jetson TX1 using NPP example mentioned in cuda samples. The time taken for encoding is 17ms and decoding is 9ms. The process takes same amount of time for me on the GPU of Jetson for encoding. Is there any way of making it faster?

You can get much faster results both for JPEG encoding and decoding with Fastvideo SDK for Tegra. More info you can get here:
That SDK has JPEG codec which is capable to compress 24-bit Full HD image (JPEG quality q=90%, subsampling 4:2:0) at less than 3 ms on Tegra TX1.