Different tools, different outputs, mainly because (probably) some default parameters cause different behaviours. I recommend the latest version of deeptools bamCoverage to have things nice and reproducible:

bamCoverage -bs 1 --normalizeUsing CPM --bam in.bam -o file_CPM.bigwig -e <int>

This will give you a high-resolution browser track at base-pair resolution (-bs 1, the lower the bin size the smoother the peaks will look, but the program runs longer and files are bigger), normalized to the count per million (CPM). You can optionally and an integer value to -e to elongate the short reads to the average fragment size (if you have that information, based on gel or Bioanalyzer).