(My) Home Network Speed Test Results: wired vs 802.11n 5GHz 300Mbps wi-fi - copper wins, hands down.

So, having acquired a new laptop, an ASUS Eee PC 1201N (Pros: dual core, bluetooth, hdmi out, 802.11n; Cons: not gigabit, not 5GHz 802.11n, takes 2x4GB memory but will only use 4GB of it, very poor battery life, $60 additional power adapters!), it seemed time to upgrade my wi-fi infrastructure too. Something gigabit with 802.11n and USB. So I also picked up a Netgear WNDR3700 Rangemax™ Dual Band Wireless-N Gigabit Router. Eventually, I discovered that the 1201n's network adapter was not gigabit, so picked up a Sabrent USB-G1000 USB 2.0 to Gigabit 10/100/1000 Ethernet Adapter Network RJ-45 Converter [I was curious, more than anything else, and figured USB 480Mbps had to be significantly faster than full duplex Ethernet 100 Mbps - NOT], and discovered that the 802.11n was 2.4GHz only (there are something like 29 2.4GHz wi-fi networks I can reach from my Dining Room, surprisingly, many of them 802.11n). So I also had to pick up a Netgear WNDA3100 RangeMax Dual Band Wireless-N USB 2.0 Adapter. The house backbone is already on gigabit switches. I've got Rogers Express for my ISP (phone lines too poor for ADSL), which is to say, 10 Mbps up / 512 Kbps down. Not that the ISP really makes any difference for our purposes here - anything internal will be far faster than out the ISP, and it's very, very, seldom that any website comes even close to the download speed limit. Ubuntu torrents do, but that is a very rare exception.

- most of the time, the 2.4GHz wi-fi will only connect at 65 Mbps. There are just too many other 2.4GHz nets around. The 5GHz always gets 300 Mbps. I'm perhaps 6' away from the router at the time of these tests.
- in some senses, OpenWRT has been a boon, in others, not so much. Although I always expected to go to OpenWRT, what drove me to it from the stock Netgear firmware, within the month, is simply that the stock firmware could not set the time - ntp is only used out the WAN port, and I'm not using the WAN port! [Not having the time set is unacceptable, let alone stupid - trying to examine log files with different times is just ... silly.] After a ridiculously large amount of reading, I traversed the OpenWRT learning curve, and OpenWRT 'just works.' [See forum threads at openwrt.org for everything you need to know for OpenWRT on the WNDR3700 - search for threads with 'WNDR3700' in the title.] If I had to do it again, today, it might take me, perhaps, 10 minutes - AFTER confirming the device was supported, and found and downloaded the firmware. Add another 10 to enable the radios and set up wi-fi. Amazing.
- the only reason OpenWRT is not a boon at the moment, is I'm not really using it for anything the stock Netgear firmware, also OpenWRT, by the way, doesn't give me. Except, of course, it gets the time. It's all just routing (to the NTP server), so why the heck Netgear has it only on the WAN port ...
- some day I'll switch DHCP, DNS, NTP, PXE, and the like, onto the router, and kill my CentOS server, but no time for that today. Given the slow speed of the router, relative to today's CPUs, that is, there may be little point to hanging a 2TB USB drive off of it, for backups and video - I don't expect it will keep up, especially to the GB. While still serving the internet. Of course, I could be wrong. Perhaps I'll find out some day.
- OpenWrt allowed me to play with channels, and use 2 channels at once [See 802.11n on Wikipedia]. In my area, there are a lot of nets on channel 11, and even more on 6. Channel 1 also has enough to be annoying. For 2.4GHz I tend to get double speed on channel 5, doubled with channel 3. Less so doubled with 9. Curiously, the built-in wi-fi of the 802.11n tends to double at 150 Mbps, while the USB adapter doubles to 130Mbps.
- Not that 2.4GHz is practical, speed wise, but with the USB adapter, and thus 5GHz, 300Mbps isn't so bad.
- Nothing, as you will see, compares with copper. Sadly.

Network speed is VERY important. I can sling GB's of backups around the home network at night. Casual tests in the past have shown that transfers go from days to hours, 100Mbps to Gigabit, full duplex both. All tests were performed under the Windows 7 that came with the 1201n, writing to the ext3 partition. Data file was a new 10GB vm disk file, except for the SD card where it was 2GB, due to the (Windows) FAT32 file size limitation. Source file is on a Windows XP machine (Pentium D, 2.80 GHz, dual-core), reading from an ext3 partition. All hard drives SATA - 300 Mbps / probably 5400 RPM source, 1201n is proably 4200 RPM.
- all results, MBps.

  •  1.08 - ISP, www.bandwidthplace.com
  •  2.32 - 802.11g, 2.4GHz, 54Mbps (linksys usb)
  •  3.34 - 802.11n, 2.4GHz, 65Mbps (1201 internal)
  •  3.46 - SD to same SD copy. {OUCH! wi-fi is faster!}
  •  3.98 - 802.11n, 2.4GHz, 130Mbps (netgear usb)
  •  4.29 - SATA to SD (1201 internal)
  •  4.79 - 802.11n, 2.4GHz, 150Mbps (1201 internal)
  •  6.08 - 802.11n, 5GHz, 300Mbps (netgear usb)
  •  7.60 - 100 Mbps wired
  •  7.90 - 1Gps wired (usb)
  •  8.72 - SATA to same SATA (1201n)
  •  9.00 - 100 Mbps wired - anomalous result, from 20 days earlier
  • 10.19 - SD to SATA (1201n)
  • 17.40 - (Source) SATA to same SATA

NOTE: Nobody should take these test results for their absolute numbers. Too many variables - these results are for my particular set of circumstances / hardware, at that particular time. I do think, however, that you can bank on the speeds relative to each technology. IMO. You could tweak any given technology to achieve better results, no doubt, but generally, for the average user, I expect one can reasonably rely on these results, relatively.

Personal Conclusions:
- since the 1201n has such poor battery life, and so I'm trailing around a power adapter anyways, and since I have wired enough jacks in my home and have sufficiently long ethernet cables, wi-fi is pointless.
- given a netbook, without a Gbps wired ethernet port, the USB gigabit adapter is pointless. The 300 KBps gained isn't worth any cost. Too many other slow technologies are present. But, one doesn't acquire a netbook for extreme speed.
- the SD results were interesting. Especially SD to itself (read & write), Sata to SD (write), and SD to Sata (read).
- 802.11n 2.4GHz is probably pointless, given the number of wi-fi homes. Go 5GHz or don't bother. CHECK THAT YOUR POTENTIAL LAPTOP PURCHASE IS DUAL BAND / 5GHz. DON'T REPLACE YOUR CURRENT WI-FI ROUTER, OTHERWISE.
- since so much of 802.11n is still 2.4GHz, no point in replacing your cordless phones, except to 1.9GHz DECT. Both keeping your 2.4GHz phones, and replacing them with 5GHz phones, is self-defeating.
- copper, particularly Gigabit, really is worth the effort. Particularly once you get over the unnecessary FUD of running cable or poking holes in your floor. Myself, I just made small holes in the corner of a room - didn't even try to go up the wall (studs), or went down the heating ducts. Haven't looked back, since. Except, to pointlessly try 802.11n, largely wasting the $.
- the results show, clearly, that 300Mbps is a fantasy. It may be true, but the associated overhead means that comparing wi-fi speeds to wired speeds is, IMO, fraud on the part of the industry. One would reasonably have thought any wi-fi over 100 Mbps would be faster than wired 100 Mbps. NOT!


Hope this is useful for anyone considering a network upgrade, particularly in wi-fi, and wondering if the speed gain will be worth the $ spent. Not likely. But my results should at least give you an in-home ballpark of what you might be looking at.

Comments welcome.


B. Jonkman wrote:
>> All tests were performed under the Windows 7 that came with the
>> 1201n, writing to the ext3 partition.
> Wait, what? ext3 on Win7 ?
>> Source file is on a Windows XP machine (Pentium D, 2.80 GHz,
>> dual-core), reading from an ext3 partition.
> What again? What manner of Windows drivers are you using? The
> only ext drivers I know of only understand ext2 at best...

The partitions may well be only ext2, lacking ext3 journaling, etc., but I don't remember for sure. i.e. Could be ext3, I don't believe journaling.
On Win XP, I have been using ext2ifs / ifs drives (http://www.fs-driver.org/), for myself for years, with little complaint. It works for me / I've been happy.
Initially I wanted larger file sizes than FAT32, and away from NTFS. And the possibility of just popping the drive/partition into a Linux system and just getting on with my day.
Having said that, there are a couple of caveats, in my experience:
- it *occasionally* gets a little bit of corruption. So you pop into Linux (dual-boot?), or fire up a LiveCD, the standard ext fs checks happen, and you go back. Never lost a file - it's usually an 'inode problem', e.g. power failure.
- hitting things really hard, repeatedly, will cause you grief over time. As a home / data file repository, it's just fine and I've never looked back. When I use it as a backup destination, with GBs of lots of (little?) files, it tends to ultimately fall over. I believe these to be windows bugs, and it feels like they occur when the driver is overwhelmed.
- this has been irritating, but survivable. Since these are backup sources or repositories, when it's really fallen apart I've reformatted the partition NTFS, restored from source or destination (each being on different machines), and gotten on with my day.
Windows does have the concept of an installable file system. So such ext drivers as this sit on top of that, in Windows, and this heavy activity, going through two non-native file systems, appears to evince a heavy traffic load bug, eventually. (Going to a native file system seems to avoid the problem, but also loses the advantages of ext in the process.) Again, I've only ever seen the problem on the write-heavy destination, not the source.
So, great for the source of the backups, perhaps not as the destination. (So put the destination on a real Linux / ext machine.)
Windows 7, or, more likely Vista, but I never did Vista to know, introduces problems and aggravations with system startup services. The software above will not load early enough to be happy. I forget for sure, but this is more than just a UAC (User Access Controls), or signed driver software issue.
For Win 7 I found http://www.ext2fsd.com/ | http://sourceforge.net/projects/ext2fsd/.
So, on my netbook, I have at least 3 partitions, and dual-boot. OS (win7), OS (kubuntu 10.04 lts), and data (Ext). Under either boot, the same DATA files and partition are used. Further, I can vm boot linux under Windows, using raw drives (for the OS). data is still accessible to the vm, as a shared (through the vm software) drive. I can also vm boot windows under Linux, but not with raw drives - the hardware changes cause significant ickiness (you need hardware profiles, and so on and so forth.) [The ultimate idea being able to live in Kubuntu, and occasionally fire up the win 7 vm - because some people insist on using MS Office.]
So, yes, ext under Windows.
(1) Pre-Vista: http://www.fs-driver.org/
(2) Win 7: http://www.ext2fsd.com/
YMMV. For me, it all just works. I have used (1) far longer, and so have more faith in it than (2), but Win 7 leaves little choice. (1) is a little more fire and forget, (2) has to run an additional icon beside the time in the lower right - essentially, the icon program satisfies the win 7 introduced startup ickiness.