Home » SBS2008 » Gigabit Ethernet vs USB 2.0 Transfer Speed Performance

Gigabit Ethernet vs USB 2.0 Transfer Speed Performance

I’ve tried finding a comparison of Gigabit Ethernet vs USB 2.0 transfer speed performance on the internet but couldn’t find anything useful as everyone compared theoretical speeds rather than actual speeds saying that Gigabit is faster.  I didn’t believe that as actual and theoretical transfer speeds are two different things.  I did the testing myself on an SBS 2008 server with a powered external Seagate USB 2.0 3.5″ 1.5TB drive plugged into the server and compared that with a Gigabit ethernet Thecus NAS and copied 5GB files from both USB and the gigabit ethernet NAS onto the server.  It’s also worth noting that the NAS is connected to the server via Cat6e cable via a quality Netgear Gigabit Ethernet Switch (with separate buffers on each channel, etc), and Both the NAS and the server have dual gigabit cards configured to run in a loadbalanced arrangement.  The result I got was 2:45 for USB 2.0 external hard drive and 3:45 for the Gigabit Ethernet NAS.  This was done on the weekend too, so there was no other load on the network.  So even though in theory Gigabit Ethernet should be faster, in practice USB 2.0 is actually faster.  So if you’re trying to work out whether to do backups to a NAS connected via gigabit lan or an external USB 2.0 drive then if performance is your main consideration the USB 2.0 drive should actually be faster.  It’s also worth noting that 2.5″ USB 2.0 drive that draw their power from USB 2.0 actually seem to run a fair bit slower than the powered USB drive.

Categories: SBS2008 Tags:
  1. Dennis
    May 11th, 2011 at 18:45 | #1

    I have a CoolMax NAS 300 gB. Hard wired cat6e. I copied a 32 gig video via NAS – 29 MINUTES! Then I tried USB – 19 minutes… That’s after trying various settings and forcing proper negotiation…

  2. May 12th, 2011 at 01:37 | #2

    Yep, that sounds about right… Performance really depends on the NAS hardware. If you want good NAS performance you need a decent NAS e.g. NETGEAR ReadyNAS with 7200RPM drives and dual gigabit ports setup to run in parallel

  3. beardy
    June 23rd, 2011 at 14:37 | #3

    I am glad I happened upon this site, I too have been disappointed with my gigabit ethernet/home server performance for backup and file transfer. I couldn’t get much above 10MByte/Sec (i.e. 100Mbit speeds) until I played with my windows settings on both boxes, forcing gigabit on,using jumbo frames.

    Copying between my machines varies, I can get up to an impressive 90/MBytes (indicated in windows) coming down from server but never more than 30MBytes back up – Same file, same drives. I have copied to the same drive, a WD 500GB 7200 in caddy, via USB2 and eSata drive, and gigabit connected by esata to my server.

    eSATA whoops them all with transfer speeds consistently over 100MByte/sec, however I find the USB2.0 performance in the same caddy about half that. So in my setup occasionally Gigabit will beat USB2.0, but not consistently.

    My WHS server is an Intel Atom based D510MO mini ITX box 1TB samsung, 0.5TB WD on eSATA). Not sure if a dual core Atom would constrain speeds, but if so why only one way?

    I tend to agree with ploogman and others, its USB2.0/eSATA if performance/time is an issue. However, you can still backup to your NAS out of hours when you’re not watching the progress bar though ;-)

  4. Joel
    July 16th, 2011 at 05:01 | #4

    I have a Thecus 4200 and used to get transfer speeds between 30 and 40 MBytes
    I recently did a firmware update and my speeds are now consistently 70-80 MBytes.
    Transferring data to external drives from either a Windows or Mac averages out at around a 1 gig/min where as from computer to Nas, I can transfer 4.5 to 5 gigs/minute.
    I can simultaneously stream 8 full HD movies (30-40Mbit bitrate) without a single dropped frame.
    Whereas with an external drive connected to my router, I get stuttering streaming just the 1 HD movie.

  5. July 16th, 2011 at 14:32 | #5

    Yes, certainly updating the firmware could make quite a difference. I havn’t tested it after updating the firmware but I currently have the latest firmware on the NAS and did the tests with the older firmware. In terms of an external drive it depends if you have it connected directly to the computer or as you do via a router. It will be significantly faster connected directly to a computer. Also when I’m talking about an external drive i’m talking about a powered 7200RPM drive (not the 5400 which gets its power from USB). I think if i did the test with NETGEAR ReadyNAS I’d get much better results.

  6. August 24th, 2011 at 11:04 | #6

    We are trying to use external drives for backup, using Acronis Advanced Backup and Recovery to an HP RDX USB drive in a Hyper-V host SBS 2011 Guest environment. The speed from the server to the USB is dismal. Around 1 hour 7 minutes for 28 GB. I believe the first post is comparing reads From the external devices to the server. Could the first poster give us results going the other way. Backup or copy the same data TO the external devices?


    Jim H

  7. August 25th, 2011 at 03:28 | #7

    That speed sounds pretty bad, so not sure what’s going on there.

  8. TX75067
    October 9th, 2011 at 15:49 | #8

    Seriously? USB 2.0 is faster than properly configured gigabit ethernet? DEFINITELY NOT TRUE! First off I’ll admit that I am a corporate IT engineer and I’ve designed and built out the network my company operates on every day. Second I’ve done some extensive testing on my home network to figure out my best storage implementation given the equipment I have available… now let’s talk about actually using gigabit properly.

    First thing is that you need a decent switch that supports jumbo frames. I’m using an older HP Procurve 1800-24G that only cost $250 brand new a few years ago. You’ll also need to make sure that the network card in your PC and the network interface on the NAS support jumbo frames. Set the frame size on the NAS to the max value and set the PC to 4k. Both network connections should autonegotiate with the switch at 1000FDX. If they aren’t then you need to check your gear.

    Once everything’s linked and set start transferring some files and see what you get. If you’re not using a NAS that’s under-powered or that has a poor raid implementation then you should easily surpass transfer rates available via USB 2.0.

    Also… notice that I didn’t mention wireless in there..? Wireless isn’t going to give you high bandwidth data transfer rates. Wireless is intended for flexible connectivity not high bandwidth data movement.

    Another thing… for anyone who’s connecting USB storage to their router… that’s not intended to be a high performance NAS. That’s intended to be a very flexible network storage option. Don’t ever expect those solutions to give you anything better than USB transfer speeds.

    Last thing to remember… the maximum possible transfer rate between any two points cannot be greater than the slowest point. SATA II is rated for 300MB/s. USB 2.0 is 48MB/s. Gigabit ethernet is 100MB/s. These are maximums… RAID implemenations that combine multiple drives on seperate SATA channels give you higher aggregate SATA transfer speeds, minus the RAID level overhead, up to a maximum of the controller’s interface rate. Best case scenario example: The drive system in your NAS can push 300MB/s of data to its own network card. That network card can push 100MB/s of data to the switch it is connected to. The switch, under optimal operation, can push that full 100MB/s across the wire to your PC. Your PC’s network card can accept that 100MB/s of data and push it to the drive system in your PC which can write it to disk in real time. If this is all true then you’ll be able to effectively transfer 100MB/s of data. If you replace the network interfaces with USB you’ve now dropped that to a maximum of 48MB/s regardless of anything else.

    The reality is that consumer gear isn’t designed for high bandwidth sustained data transfer. Everything from your drive system to your network card is depending on your CPU to make things happen. For machines with hardware RAID controllers and “server grade” network cards the case is much different since both devices have their own processors and the RAID cards have dedicated onboard memory. On the NAS side you’re facing the same type of issue unless you’ve spent some nice coin on a higher end machine. Many home office NAS boxes today have better specs but also have a crap load of “apps” that’ll do everything from streaming media to serving webpages and downloading torrents. Every additional feature you enable will require processor time, memory space, and will take up at least a small amount of drive activity. Keep all of this in mind when you start troubleshooting or optimizing performance. Know your gear and how it is used. Always figure out what the slowest link is between the systems and realize that you’ll never optimize past that point.

    I know this was long… I hope it helps…

  9. TX75067
    October 9th, 2011 at 15:54 | #9

    I really just can’t wrap my brain around how this guy can actually post this online and feel that it is truly correct… astounding…

    Your test doesn’t mean that the theoretical rates are wrong… it means that you have a problem that needs to be fixed. Any network engineer that is worth two pennies can prove this to be completely wrong in no time.

    I’m sorry… I just hate it when people present themselves as sources of technical knowledge and provide completely wrong information with nothing to back it up. Other people who are looking for help will stumble across this and think, “hey I don’t have a problem with my setup because this guy blogged that it should be slower”… dude I’m sorry but this is crap.

  10. October 10th, 2011 at 01:29 | #10

    Very good and insightful comment, thank you! One problem that people seem to forget (including professional network engineers) is that there’s a difference between theoretical throughput and actual throughput. In particular this tends to present itself when dealing with networks. You have to remember that the NAS isn’t the only thing communicating with your server and your switch has to deal with up to 48 other connections as well. There is also the issue of interference and packets being dropped, which also slows things down. With USB these problems tend not to manifest themselves as much. As I have said, I’m of the view that the crappy throughput was in part because of the NAS, which was a Thecus 4100 PRO. Perhaps with a high-end Netgear ReadyNAS or a QNAP you will be able to achieve better performance. I will also check jumbo frames. From memory, I actually turned these off as we had issues with a Repotec switch we were using, but I don’t think it will be a problem with a high-end Netgear switch we’re using now. I still think in the end the result will be very similar to what’s achievable with USB (perhaps a little better), because of all the other traffic on the NIC on the server (which btw I had trouble configuring into a dual-NIC as SBS2008 doesn’t seem to like the driver and actually bluescreens).

  11. JP
    November 10th, 2011 at 15:45 | #11

    Can you run the test again and plug the NAS directly into a network card on your server discounting the ethernet switch. What are the results now?

  12. November 17th, 2011 at 07:56 | #12

    That’s probably not a bad idea but not something I’m going to do as for me it’s not a realistic scenario. Worth doing though if you want to just test what’s going on. I double it would be the switch on my end as I’m using a nice Netgear 48 port gigabit switch. I think enabling jumbo frames would help though.

  13. Steve
    December 23rd, 2011 at 01:25 | #13

    -and back to USB testing speeds!

    I had done a USB 2.0 speed test for tethering internet from my cellphone’s internet access.
    The USB 2.0 cable is an HTC phone adapter USB cable.
    When I had connected and tethered the internet I looked at the status of my connection and it said that i was running an internet speed of 425.9 Mbps.
    Perhaps we will be using newer USB cables for internet rather than that limiting Ethernet cable with only 100 Mbps or 802.11 wireless with only up to 54 Mbps.
    My suggestion would be to use a USB for future devices like from modem to router. or just newer wires that have a faster transferring speed.

  14. December 27th, 2011 at 02:34 | #14

    I think this is a slightly different topic to your comment. If you’re tethering to your cellphone for internet access I don’t think the connect to the phone will be an issue… There’s no way your phone would get you 425.9 Mbps internet access…. That’s WAY too quick for any 3G wireless.

  15. January 19th, 2012 at 12:13 | #15

    Hi All,

    Just had to post this – as I can believe what I’m reading…USB2 is *not* faster than Gigabit Ethernet – unless your hardware / software setup is seriously incorrect.

    I am a technical support engineer from a corporate environment, as I write this I am transferring 37GB of data from one (cheap) PC to another over GigE – just SATA300 disks no RAID, no fancy controllers, a single NIC and really cheap GigE switches and CAT5e cables.

    It took 12 minutes…..that’s 37GB of data in *12* mins….might be worth looking at that old PC before you splurge for a Drobo :-)

    If this was USB2, I would have time to go and have lunch before it would finish :-)

    Basically TX75067 has it bang on – including the fact that wireless is not used for speed, but connectivity – and it’s a similar story with USB2: fantastic, easy to use connectivity – but give me GigE any day for speed.

    If you need fast (cheap) backup, never use a basic NAS – an old PC will blow it away. In the test above – the PC I’m copying *to*, is 6 years old, albeit with nice new hard disks.

    Just a point from experience – the slowest link is almost always the hard disks write speeds – unless you can afford SSD or really good RAID storage. The disks I used in the test above can only manage about 60MB/s flat out – then you can pile on top the TCP/IP overheads and any software used – backup programs typically have to compress and encrypt, which all sours the deal.

    Hope this helps.

  16. January 19th, 2012 at 21:27 | #16

    You’re right in that you really need to test your environment. As I pointed out we were doing this with a Thecus NAS, not a PC. The problem with NASes (especially not the top of the line ones) is that they have pretty crappy CPUs and not that much RAM, which really slows things down. Also as the disk speed is often a limiting factor the RAID setup matters. It’s far better to have a RAID 5 or at least a RAID 10 rather than a RAID 1 only because with RAID 5 and RAID 10 you are striping across multiple disks thereby increasing IO throughput. SSDs aren’t really an option for backups quite yet because of the cost. You’re right that an old PC will most likely have far more grunt than any NAS and you can install FreeNAS on it without too much trouble. I personally really like the HP N36L or N40L MicroServers as they’re the size of a NAS and take 4 bays with 1 extra instead of the DVD and you can run the OS of an internal USB slot and all for between $200 and $300 with a FreeNAS setup with z-raid. Jumbo frame support on your network can also make quite a lot of difference, as it can signficantly increase throughput when transferring large amounts of data.

  17. March 14th, 2012 at 18:58 | #17

    Definitely gigabit over USB 2.0. We get asked that question pretty frequently when customers try to determine ROI of a new cabling system. You need to invest into a decent equipment capable of handling jumbo frames, make sure that the MTU of your network nodes matches and use quality cable.
    Of course, USB 3.0 is a different story.

  18. March 18th, 2012 at 13:33 | #18

    Hi, The performance of Gigabit Lan is much better in comparing to USB 2.0

    i did a test on Gigabit Lan:
    – Buffalo NAS, 2x500GB HDD on RAID
    – Gigabit Switch, some name of china
    – Macbook, Windows OS, Teracopy software

    I repeated the test in working hours since other users are connecting to the NAS to read/write documents:
    – Copy from NAS to internal HDD: 40MB/s, which means ~2.1 mins on transferring 5GB file
    – Copy 2 files from NAS to internal HDD and another external USB HDD parallely: ~60MB/s
    which means ~1.6 mins on transferring 5GB file.

    For today, i would suggest:
    – SSD with USB3.0 box if the server has USB3.0;
    – Or a NAS with 2 or more HDDs in raid;
    – Or support SATA3 with SSD. The reason of Sata3 is because so many old model of NAS is low performance.

  19. Ty
    March 25th, 2012 at 04:40 | #19

    It’s even worse than the case you presented, newer old hardware supports tons of new interfaces but they seriously skimp on overall throughput. I got ~60MB/s to IDEs and ~70MB/s to SCSI array over Gb with a single nic and a cheap dynex switch. Best I can get from a sata + 3.4 ghz dual core is ~50MB/s, though it usually caps at 25MB/s, varies each time it’s connected. The only reasons I switched were to save about 10-15 bucks a month in electric (SCSIs were 10k RPM) and the dual PII xeon was prone to packet flooding as it also served firewall and VOIP duty, locking up or forcing a reboot 2-3x a week. Systems between then and now can still get good throughput but their handling of multiple data streams sucks.
    ICH6 and under are kind of useless as they choke easily on multiple IO of any magnitude, their saving grace is they don’t crash the OS. The old 700mhz thunderbird could process 3 video streams concurrently for 40-50MB/s over the network, burning a DVD at 16x while doing internet browsing! In newer systems 12x is a challenge if much of any operations are going simultaneously, this goes for AMD and Intel. What’s all that processor speed get if data doesn’t have anywhere to go?

  20. March 26th, 2012 at 14:54 | #20

    Thanks a lot for the real world testing and figures. I have an old Compaq laptop with Pentium Dual 1.73GHz and 2GB RAM. I have connected an internal Seagate 1TB HDD to a Piranah (generic) USB/eSATA enclosure. I only get around 23MBPS while writing to the HDD.

    One thing wanted to point out is that, USB is length limited. So if its long distance …. Still I got your point.

Comment pages
  1. March 28th, 2011 at 23:51 | #1

8 − seven =