Return-path: Received: from lo.gmane.org ([80.91.229.12]:39941 "EHLO lo.gmane.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752406Ab0ENUNA (ORCPT ); Fri, 14 May 2010 16:13:00 -0400 Received: from list by lo.gmane.org with local (Exim 4.69) (envelope-from ) id 1OD1Ff-0004dl-VG for linux-wireless@vger.kernel.org; Fri, 14 May 2010 22:12:56 +0200 Received: from CPE0014bf4aedf8-CM00111ae2cca2.cpe.net.cable.rogers.com ([99.234.24.234]) by main.gmane.org with esmtp (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Fri, 14 May 2010 22:12:55 +0200 Received: from senada by CPE0014bf4aedf8-CM00111ae2cca2.cpe.net.cable.rogers.com with local (Gmexim 0.1 (Debian)) id 1AlnuQ-0007hv-00 for ; Fri, 14 May 2010 22:12:55 +0200 To: linux-wireless@vger.kernel.org From: seno Subject: Re: RTL 8187b - naughty behaviour Date: Fri, 14 May 2010 20:12:44 +0000 (UTC) Message-ID: References: <4BB7660A.3030900@lwfinger.net> <4BE30879.4020008@lwfinger.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-wireless-owner@vger.kernel.org List-ID: Larry Finger writes: > I do not understand the poor performance. I have a device with the > rtl8187b chip - USB IDs 0bda:8187. That is an ID for an rtl8187, but it > does have the B chip. I use tcpperf to measure the transmit speed. At 2 > m from the AP with an indicated signal of -17 dBm, my throughput is 13 > Mb/s. At 10 m with a signal of -49 dBm, the transmit speed is 3.5 Mb/s. > On the download using speedtest.net, I got 11.3 Mb/s at 2 m, and 2.64 at > 10 m. I got 0.49 Mb/s upload at both distances as that is the speed of > my line. > > What happens if you use iwconfig to force a particular rate? At what > setting is your throughput maximized. I don't think there is a problem > with the rate setting mechanism, but that would test it. > > Larry Hello, it turns out that terms 'poor' or 'slow' performance on 8187b chipsets need to be discussed more in detail. If I would have never had a faster network, I would possibly not notice those performance downturns. I did some test on my notebook, once with mandriva 2010.0 and once with win 7. The results: windows: @ 2 meters distance rate : 54M dBm : -46 speedtest: 9.9/0.98 Mbit/s (down/up) DL file*: ~2.5 MByte/s @ 10 meters distance rate : 54M dBm : -64 speedtest: 9.9/0.97 Mbit/s (down/up) DL file*: ~1.25 MByte/s linux: @ 2 meters distance rate : 36M dBm : -30 speedtest: 9.9/1.00 Mbit/s (down/up) DL file*: ~1.5 MByte/s tcpperf output: 11769623 b/s (11769.62 kb/s 11.77 Mbit/s) @ 10 meters distance rate : 18M dBm : -45 speedtest: 1.2/0.85 Mbit/s (down/up) Dl file*: ~130 KByte/s (<---!!) tcpperf output: 3663403 b/s (3663.40 kb/s 3.66 Mbit/s) Dl file* = copy large file from a machine which is connected via 100mbit ethernet to the router. I did all tests several times to make sure to not report some random results. Before I used pid algorithm as someone described above, but could not notice any difference in network behavior. If rate is set to auto, it _sometimes_ (~ twice a day) happens that connection is up and iwconfig shows a rate of 24M, but I can not even ping the router. Same happens if I set a fixed rate of 36 or 18, but it happens more often (every 15 min) that the connection is up but no traffic is possible at all. A fixed rate of 5.5M gives the most stable connection although the wireless speed is as slow @ 10 meters with auto rate (130 KByte/s is FAR from 5.5M) The rate settings apply to 10 meters distance, which is my typical situation here at home. On another PC with zw1211 wireless I set rate to 54M some time ago and there the trick did it, connection is perfect and very fast.