after a while I got to do some benchmarks with an Archer C7 with a hw version 2 card I received from QCA
with latest trunk (with backports of 20131105)
(btw trunk before that on a WDR3600 had problems: I couldn't associate the Archer C7 to that unit)
and with http://patchwork.openwrt.org/patch/4396/
(running with haveged because entropy_pool was really low)
in preparation of later testing with AC modes i use a custom hostapd.conf file (and wpa_supplicant package)
interface=ath0
driver=nl80211
ssid=ath10k
channel=40
country_code=US
ieee80211n=1
ieee80211d=1
ieee80211h=1
ht_capab=[HT40-][SHORT-GI-20][SHORT-GI-40][DSSS_CCK-40]
hw_mode=a
macaddr_acl=0
logger_syslog=-1
logger_syslog_level=0
logger_stdout=-1
logger_stdout_level=4
wmm_enabled=1
it seems to work:
iw dev ath0 info
Interface ath0
ifindex 548
wdev 0x203
addr 00:03:7f:48:da:38
ssid ath10k
type AP
wiphy 0
channel 36 (5180 MHz), width: 40 MHz, center1: 5190 MHz
Results with iperf: ath10k
Archer C7 -> WLAN -> Client (ath9k_htc card: WNDA3200 2x2 card so only "300Mbit/s" speed)
[ 4] 0.0-300.0 sec 2.14 GBytes 61.3 Mbits/sec (max: 83Mbit/s ; min: 1Mbit/s)
compared to Archer -> LAN -> Client
[ 4] 0.0-100.0 sec 3.89 GBytes 334 Mbits/sec
iperf-PC -> LAN -> WDR3600 -> WLAN (2x2 @ 2.4GHz highly populated) -> Archer C7 -> WLAN (2x2 @5GHz) -> PC
avg 25.7 Mbit/s
During the test there were periods lasting 30 seconds where the transfer rate broke down to 1Mbit/s from the "max" of >70Mbit/s
However the avg. transfer rates seem to be normal compared to a 2x2 setup: http://www.smallnetbuilder.com/lanwan/router-charts/bar/113-5-ghz-dn-c
So I think these 30sec "windows" are only some artifacts / protocol inherent