Performance varying without and without output dumping

Hi friends,
  We are facing some problem while testing video codec on pandaboard.
Problem statement:-
In case of a video decoder we have timer code  to take the fps
reading.The fps reading is coming different with and without output
dumping(yuv write).

Its like :-

  1st scenario:

  clk_begin();
 decode_fuction();
  clk_end();
  dump_outpt();

 2nd scenario:
clk_begin();
 decode_fuction();
  clk_end();
//  dump_outpt();  (dumping yuv output is commented)

In case of 2 nd scenario we are getting 2 more fps than 1 st
scenario.
The dumping function is performed after collecting the total time for
the decoding function in case of 1st scenario, so ideally it should
not have any impact on the timer output.
 Please suggest where it is going wrong or what the specific reason
for drop in fps when dumping the output(fwrite for yuv dumping) ,
 even though we are dumping the yuv after the timer gets the total