Forget to mention the comparison of quality of normalization/warp between SPM & afni:
In SPM, warping EPI images to TT_EPI+tlrc template ( come from afni and change into nii format): when use fsl bet2 to remove the skull, EPI alignment with EPI template is very good.
In afni, use @auto_tlrc script to warp EPI images to TT_EPI+tlrc:
with its own skull strip or using fsl bet2 skull-striping and no_ss option, the alignment is OK with obvious mismatch at occipital and frontal regions.

Maybe the way I am using afni to warp EPI is not correct. But I also try the old way (warp T1 anatomy to T1 template and then adwarp EPI images, does work very well either due to image distortion & subject motion). Hope someone have a better solution.