DVRRDK question on processors loading

The platform I am using is DVRRDK 3.0 and 816x DVR.

If I have a chain of  capture -> nsf -> dei -> dsp -> smms -> display, 
                                                                    |------------------|  
the CPU loading of m3vpss, m3video and c6xdsp are about 40%, 1% and 15% 

but if I modify it to: capture -> nsf -> dei ->  enc,
                                                                  |--------------> swms -> display
 the CPU loading becomes m3vpss, m3video and c6xdsp are about 70%, 50% and  1% 

and if I add the dsp osd back,  capture -> nsf -> dei -> dsp -> enc,
                                                                                        |--------------> swms -> display
 the CPU loading becomes m3vpss, m3video and c6xdsp are about  87%, 50% and  40%

The output from DEI to DSP or encoder is 16 ch 720x576 and 360x288 at 25 fps
is the serious increase in loading of m3vpss and c6xdsp normal?
or is there is something that I have set wrongly causing this? 

Thanks