So you guys inspired me to do a bit of testing. It seems Samba at least is not at fault:

Code:
root@Mars:/mnt/user/store/temp# dd count=1048576 if=/dev/zero of=/mnt/store/temp/asdf
1048576+0 records in
1048576+0 records out
536870912 bytes (537 MB) copied, 34.3965 s, 15.6 MB/s

root@Mars:/mnt/user/store/temp# dd count=512 bs=1048576 if=/dev/zero of=/mnt/store/temp/asdf
512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 21.0785 s, 25.5 MB/s


Interestingly during the small blocksize test dd used about 25% CPU and shfs about 70%. During the 1M blocksize test, shfs sat around 10-12% and dd around 3 or 4%. So

Code:
root@Mars:/mnt/user/store/temp# time dd count=1048576 if=/dev/zero of=asdf
1048576+0 records in
1048576+0 records out
536870912 bytes (537 MB) copied, 35.7917 s, 15.0 MB/s

real    0m36.158s
user    0m1.360s
sys     0m6.570s

root@Mars:/mnt/user/store/temp# time dd count=512 bs=1048576 if=/dev/zero of=asdf
512+0 records in
512+0 records out
536870912 bytes (537 MB) copied, 21.2011 s, 25.3 MB/s

real    0m21.551s
user    0m0.010s
sys     0m0.790s
_________________________
Christian
#40104192 120Gb (no longer in my E36 M3, won't fit the E46 M3)