Dear Developers,

during the last weeks i've been experimenting with multiple AR9x Atheros cards 
driven by ath9k.
I did some 802.11n throuput tests with an Ubiquiti SR-71-12 (AR9220-AC1E) and 
Unex DNMA-H5 (AR9160-BC1B) in an Alix 3D Geode Board and another Ubiquiti 
SR-71 (AR9220) in an Acer Dual-Core Atom Netbook.
I've been using linux-next checked out on Friday, 06.09.2010 for all my tests.

First, I measured the performance against the cards themselves. I always got 
around 90mbit/s, HT20 AND HT40. Then, i connected to an Apple (don't know 
which card), Win7 Intel 5100, and a D-Link DIR-825 Router (AR9220, Atheros 
Blob). The results always were just the same: Sending rate towards these 
devices was always around 20mbit/s, MCS varying from 3 to 5, receiving around 
90mbit/s. I did all my tests in the same room.
There was another network at g band, but not a. But results were the same.
Using adhoc, AP or station with iw or wpa_supplicant also made no differences.

To exclude environmental influences, i connected the Atheros cards to the D-
Link by wire. ath9k was sending with full MCS15 SGI, 240mbit/s from the 
netbook, 220mbit/s from the Alix board (iperf UDP). Looks fine.
I tried ath9k_rate_control and minstrel_ht with the same results.
Are there problems with signal quality calculation or what forces the rate 
algorithms to choose such a low bitrate?

Rx still didn't exceed those 90mbit/s. As the Netbook operated at the same 
rate, I think I can exclude CPU problems.

There must be some reason limiting the overall transfer rate to ~90mbit/s!

Examining debugfs a little, i can see there are no rx drops nor lost 
interrupts. Does anyone have a hint what this could be related to? Where 
should i start investigating?
I have experience with driver programming, but ath9k and mac80211 are still 
pretty new to me.

Thanks in Advance,

Alex
_______________________________________________
ath9k-devel mailing list
[email protected]
https://lists.ath9k.org/mailman/listinfo/ath9k-devel

Reply via email to