WiFi reception differs on Windows and Linux

[mgw]

I have 2 similar T60's at hand. One is running XP, the other Ubuntu. The Ubuntu machine reports 25% signal strength, the Windows machine reports 82% - both at the exactly same location.

Appreciate any idea what causes this issue... Txs

[Marin85]

I had basically the same experience with Intel Pro 3945abg card. But I have the exact opposite experience with the Atheros ThinkPad abg card (partly because for some reasons most of the Vista driver versions for this card make it "weaker" for some reason ).

[mgw]

Both are running on Atheros Communications Inc. AR5212 802.11abg NIC

[Volker]

The reported percentage is essentially fictional, and cannot be compared between different drivers and/or hardware. Its standard practice to "optimize" drivers to report higher numbers as it makes the customers happy. Same with the "bars" on cell phones... there have been firmware updates for phones that only change the number of bars reported

[mgw]

The reported percentage is essentially fictional, and cannot be compared between different drivers and/or hardware. Its standard practice to "optimize" drivers to report higher numbers as it makes the customers happy. Same with the "bars" on cell phones... there have been firmware updates for phones that only change the number of bars reported

Point taken but the machines actually behave exactly that way - i.e. if I take both to another room the Ubuntu machine looses contact

[Volker]

There are two different linux drivers for AR5212, madwifi and atk5k. Madwifi is older, but probably still more stable. In any case, try the other one. Updating to the newest version probably helps, too. Which version of ubuntu are you using?

[mgw]

There are two different linux drivers for AR5212, madwifi and atk5k. Madwifi is older, but probably still more stable. In any case, try the other one. Updating to the newest version probably helps, too. Which version of ubuntu are you using?

Linux michael 2.6.27-11-generic #1 SMP Thu Jan 29 19:24:39 UTC 2009 i686 GNU/Linux

8.10 that is

[Volker]

Ubuntu 8.10 is afaik the first version using ath5k by default, and that version is pretty new. Try the most recent ath5k (upgrade to newest kernel or get compat-wireless from http://wireless.kernel.org) and/or the old madwifi (http://madwifi-project.org) driver.

[mgw]

Ubuntu 8.10 is afaik the first version using ath5k by default, and that version is pretty new. Try the most recent ath5k (upgrade to newest kernel or get compat-wireless from http://wireless.kernel.org) and/or the old madwifi (http://madwifi-project.org) driver.

Just downloaded the latest code from
http://wireless.kernel.org/en/users/Download
the instructions on the site are pretty much comprehensive and can be performed on the fly without rebooting - the only thing missing is the last statement to load the wireless service again:

Code: Select all

sudo make load ath9k

WiFI reception increased from 45% to 75%.

[bluesceada]

You also can't compare two different machines... One might just have a better connected antenna than the other, these can be quite sensible.

[mgw]

You also can't compare two different machines... One might just have a better connected antenna than the other, these can be quite sensible.

As stated in the initial post: 2 identical machines

[bluesceada]

As stated in the initial post: 2 identical machines

The SAME machine with one time Ubuntu as the OS and one time XP, or two theoretical identical machines? No machine is identical to another, there are always variances in the production....

(The same with the location by the way, depends on monitor tilt etc., a milimeter at a different position does count)

[mgw]

The SAME machine with one time Ubuntu as the OS and one time XP, or two theoretical identical machines? No machine is identical to another, there are always variances in the production....

(The same with the location by the way, depends on monitor tilt etc., a milimeter at a different position does count)

I went through this is detail and yes, there is a hughe difference in reception between my Linux machines at the same spot with the same angle/tilt of the screen... by the way: in today's production environment no machine would leave any plant with deviations from a production standard like this.
In essence, all our machines have a 100% reception in that spot, provided they are booted with Windos. It's a known problem with the Atheros drivers (not the chips). It gets better with ATH9K, though.

[bluesceada]

Yeh the differences between 45% and 82% are too huge.
But between 82% and 75% it is quite possible. Especially wifi-antenna connections...

[mgw]

Yeh the differences between 45% and 82% are too huge.
But between 82% and 75% it is quite possible. Especially wifi-antenna connections...

Agreed - but keep in mind that the initial difference (on another floor of the building) has been 25% vs 82%

[GomJabbar]

It is possible that you have a disconnected antenna wire in the Linux machine. You could rule that out by swapping hard drives.

I know that Windows has power settings for the WiFi cards. Not sure if there are such settings in a Linux GUI, but there may be command line switches to change the power setting of the card.

[gongo2k1]

My experience has been that linux wifi drivers handle power saving modes differently (which affect the actual signal strength).
The percentage numbers that either side reports are largely fiction since the drivers don't take noise into account.
You can get good signal even from far away if there's no noise, and it'll report very low reception, but in another situation you could get almost no signal despite being only 50 ft away if there is a lot of noise on that band (and windows will still report excellent reception).

Also, if you're running Ubuntu, you could always run the live cd, no need to install/dual-boot/swap hdd. Then you'd have a straight apples-to-apples comparison.