Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

The systems:

Del Optiplex w/ 1GB eth card , win7

mini-itx system, jetway nf98, intel i3,1 GB onboard eth card, crux w/ custom

The network: cat 5e or 6 all around. systems are less than a foot from each other. longest cable is 6 feet. have tried variations of router only, router and switch, switch, or direct connect (no router, or switch) (switch & router are gig too btw)

the share:

i mount the crux system on the windows system using samba

i mount the windows share on the crux system using 'mount -t cifs'

NOTE: variations of only one mount at a time, both mounts at a time, prob anything you can think of

Transfer 1: while using the win7 system, i start>run>\\share and up pops an explorer box displaying the contents of my internal drive on the crux system. i then drag a 10g file from the share to the desktop of the win7 system. 10g file was made 'dd if=/dev/sda of=/a/10g.dd bs=1M count=10240' THIS results in a transfer at speeds: 55mb/s (lowest) to 105mb/s (highest) and avgerages about 75mb/s. [these results are above goal]

Transfer 2: while using the crux system, i mount the win7 share ('mount -t cifs-o //192..../share /a'), and ls -l /a displays contents of win7 drive's share. i then time a copy of a 10g file from the share to the internal drive of the crux system. THIS results in transfer speeds: 5m23s so ~ 31 mb/s

why are the results so different?

the highest i can set mtu on crux w/ this card is 4000, but that seems to have affect of zero.

i also try dd, using bigger block sizes, which gives little improvement

EDIT: It occurred to me that it might be the windows7 copy is faster/better. SO, from the windows machine, in cygwin, i executed 'time cp //192.../share/10g.dd /cygdrive/c/../Desktop Resulted in 2m51s ~ 59mb/s

share|improve this question

migrated from Dec 7 '11 at 20:07

This question came from our site for professional and enthusiast programmers.

The title and text keep on mentioning "network transfer" and "transfer". Your tests are more timings of a file copy rather than "network transfer speed", and the bottleneck is probably at the file creation/write rather than the "transfer". What happens if you eliminate the network, and just perform local disk-to-disk copies? This would provide a reference time for each filesystem, that is, ntfs and ext4(?). – sawdust Dec 7 '11 at 22:24
You should also try reversing the direction of the file copies, or what you call "transfers 1 and 2". – sawdust Dec 7 '11 at 22:59

What are the specifications of every card (chipset, etc).

There can be huge difference in performance between cards. As an example, server / enterprise grade cards have their own circuits to perform tasks that in basic cards are delegated to the main processor.

So, may be, it is just a difference in hardware performance.

share|improve this answer
crux system's nf98 motherboard (intel QM57 @ 3.5W) onboard lan: Intel 82577LM Gigabit Ethernet optiplex 380 ( Intel® G41 Express Chipset with ICH7): Broadcom® Integrated Gigabit Networking (BRCM 57780) – c card Dec 7 '11 at 20:56
@jap1968, if you need to ask a question, it isn't an answer, please use the comments fields to get more information in order to provide a complete anaswer. – Paul Dec 7 '11 at 23:06
@Paul: Ok. My mistake – jap1968 Dec 8 '11 at 18:10

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.