WMM limiting download throughput

Discussion in 'Tomato Firmware' started by helkaluin, Aug 27, 2012.

  1. helkaluin

    helkaluin Networkin' Nut Member

    Router: Asus RT-N12 C1 (identified as B1 on Tomato web admin page)
    Tomato firmware mod/version: Shibby mod v.99 (tomato-K26-1.28.RT-N5x-MIPSR2-099-Max)

    No other traffic is present during testing. QoS is disabled.

    With WMM:
    Without WMM:

    Is this intended behaviour? I know that without WMM I would be limited to G speeds. But enabling it (and thus enabling N speeds) seems to have a counter-intuitive impact on my download speed.

    UPDATE: Flashed Shibby v.100, cleared NVRAM, and WMM is still giving me the same impact on download throughput.
  2. helkaluin

    helkaluin Networkin' Nut Member

    Uh, could anyone knowledgeable at least reply to say if the observed effects of WMM is intended?

    Real life browsing (that is, not using synthetic benchmarks like Speedtest) shows that flipping on WMM here does cause slower page loads, especially when large images slowly appear line by line.

    If anyone has an RT-N12 themselves to test this then all the merrier. But if you don't have the exact model, (as I should assume,) then perhaps at least throw in an educated guess on the theoretical matters concerning WMM's impact on throughput?

    Enabling WMM is the only way to get N speeds as far as I've read. But it seems this choice is making my general internet usage go slower!
  3. helkaluin

    helkaluin Networkin' Nut Member

    I've flashed back the latest stock firmware, and the results are similar.

    With WMM:

    Without WMM:

    This raises the question: could the impact be affected by hardware? One possibility is that the stock firmware and Shibby Mod shares some code regarding WMM behaviour, but I have no idea about that. If not, than it's likely that I have a faulty model. Or may be I don't, if only someone can stand up and say something about this throughput test is actually intended behaviour.
  4. pharma

    pharma Network Guru Member

    You might also try setting all Advanced Wireless settings to defaults and testing, then change only WMM and test. Trial & error with other wireless settings you'll notice other settings will impact your wireless effectiveness. If thing aren't still working correctly, try an older version of Shibby's firmware -- one that is not using the newer broadcom driver. I'm using Toastman's firmware (see signature) and don't experience this issue using his older .0500.2 firmware.
  5. helkaluin

    helkaluin Networkin' Nut Member

    I've already been testing with defaults (NVRAM reset) and dissecting combinations of different settings. None effect the drastic throughput change as shown above with flipping WMM on and off.

    I've now downgraded to Shibby v.92, just before the driver is committed. And wonderfully the throughput is no longer affected negatively by flipping WMM on.

    So this is the Broadcom driver playing bonkers with the BCM5357 SoC wireless NIC. Is the driver GPL and in the Linux tree or is it proprietary? Perhaps I should file a bug report, if one doesn't exist already.
  6. StarClout

    StarClout Serious Server Member

    I'm gonna test this on my E2500. Been having some fluctuation in speed (20/2 Cable connection) so I'll report back
  7. helkaluin

    helkaluin Networkin' Nut Member

    Damn. It seems I was too quick to put forth conclusions.

    Now with Shibby v.92 and WMM on the download throughput issue pops up again.
  8. Toxic

    Toxic Administrator Staff Member

    I have seen this type of post before. It is a misconception that your speed will be the same. This is QoS for essential serves for voice and media. It does not guarantee throughput.
    If you are just downloading files. Then turn WMM off
    Sent from my GT9100 - CM10 - Tapatalk v2.2.8
  9. pharma

    pharma Network Guru Member

    Exactly what Advanced wireless changes are you making? Are you just changing the WMM with everything else at default?
    Toxic is probably right that there won't be a guarantee of same speeds when it is turned on. When it it turned off do you get N speeds? Because you might not need it turned on and is fairly easy to test.
  10. helkaluin

    helkaluin Networkin' Nut Member

    I've been leaving everything as default. I've been even leaving it as an open network. Setting it to WPA2-AES doesn't change anything. The moment I thought flipping on the UPnP daemon caused the trouble, neither disabling it afterwards nor resetting NVRAM gave back a reasonable >1Mbps download throughput.

    This is really causing more trouble than it's worth. I'm disabling WMM now even though it's clearly restricting the theoretical speed to G limitations. (Looping iwconfig shows a maximum of 54Mbps.)
  11. Elfew

    Elfew Network Guru Member

    I have no problem with that on ASUS RT-16n Shibby AIO v100

    speed with/without WMM are almost same
  12. helkaluin

    helkaluin Networkin' Nut Member

    RT-N16 runs with BCM4718+BCM4329 + BCM2050

    RT-N12 C1 runs with BCM5357
  13. Stefano J. Attardi

    Stefano J. Attardi Serious Server Member

    FWIW, I have the same problem with both an Asus RT-N16 and a Cisco E1000. Whenever I turn WMM on, 802.11n speeds inevitably slow down to a crawl. I've tried recent Shibby and TomatoUSB builds.

    This nasty bug has been bothering me for months and I still haven't found a workaround. I need WMM on otherwise my iOS devices won't connect. The only workaround I have is to switch back to G mode, which is sad because it limits my speeds at about 1/4th of what my cable connection can do. :(
  14. Doolick

    Doolick Reformed Router Member

    struggling with this issue on a Cisco AP1042N (enterprise-class) access point. For Internet, I am seeing 10-14mbit speedtests instead of 16-17mbit with no WMM. On the LAN side, however, if I transfer a 200MB file over wireless, it goes at 3MB/s with WMM off and 8MB/s with WMM on. So, if the wireless is connecting business users to file and application servers, WMM would probably be a HUGE benefit by increasing that throughput. It would seem as though the multi-stream and multi-band features go away with WMM off, and you're left with a simple 54Mbps connection, although seemingly better than straight G, it's still limited at that speed.

    From what I can gather on the way it works, the degradation in performance is because in the wireless realm, WMM provides QoS by adding a natural delay to the low-priority packets. This is how it avoids needing a fast processor. While it seems to work, the unfortunate side-effect is that it slows down the internet (background) traffic even when there is no higher-priority traffic present. Also, turning off WMM seems to disable the multi-channel N features?

    For reference, my laptop currently uses an Intel Centrino 1030 wireless card (single-band N) and I was troubleshooting a Cisco AP1042N (dual-band N) - comparing live against 2Wire Uverse modem (single-band G) and an Apple Time Capsule (dual-band N). I also tested with a different laptop which had a Centrino 6250 dual-band card. With WMM on, I get 144Mbps and 300Mbps connections, respectively. With WMM off, speed drops to 54Mbps for both laptops. File transfers confirm this speed change at 8MB/s and 15MB/s respectively, or 3MB/s with WMM off. The math doesn't exactly work out here (bits-to-bytes) but the max transfer speed seems to be around 50% of the rated speed in all cases. Also, on the Internet performance, the laptop with the dual-band card was able to hit the full 17+ Mbit download speed with WMM on. I'm guessing this would top out around 25-30, if your connection goes that fast.

    I think the overall internet performance with multiple wireless clients will be better with WMM on, but I could be wrong about that.
  15. Victek

    Victek Network Guru Member

    Unfortunately Intel Wireless card and Broadcom are not good friends, in fact they are competitors... then one image is better than...

    (Same testing conditions)

    RT-N16 with Intel 6300 agn wireless card ... transfer WAN-WLAN (Wired rate is 100/10)

    RT-N16 with a cheap USB adaptor TL-WN821Nv2 (12€) with Realtek chipset test...

    But Intel cards don't play so good with Openwrt in Attitude 12rc2 ...

    So, first, Tomato is good firmware, and second .. may be you need to replace if possible Intel wireless card ... I hope you know the solution now...
  16. Toastman

    Toastman Super Moderator Staff Member Member

    In full agreement with Victek. I have many thousands of clients, and any time I have a complaint from one of them it turns out to be Intel or Apple wireless. At one of the local universities, many people using Intel wireless in their laptops have huge problems connecting to the campus AP's. Many of them simply have to change their laptops for one that works or use a USB adapter. Even the cheapest chinese usb adapter works better than Intel. If you google this, you will find Intel are not very popular.

    You may find some relief by tinkering with the power saving settings in the Intel setup, if it has any. Turn off any "feature" that looks suspicious.
  17. Elfew

    Elfew Network Guru Member

    I had problems with my intel 1000bgn... Today I have bought tp-link tl 822n.

    Now I can enjoy really high speed over wifi... About 50% higher than before. I recommend it to everyone who has problems with transfer rates
  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice