WNDR3700 wireless speed issue

[redcow]

Hi,

I bought a WNDR3700 netgear router and a mini PCIe Intel Wifi Link 5300 wireless card. My wireless speed is terrible I tried with the original netgear firmware and with openwrt its nearly the same, iperf gives me 35-40 mbits being next(!!) to the router and 100% signal, windows wireless connections shows me 270-300Mbps and in the Intel wireless settings I enabled for both 2,4ghz and 5ghz auto as speed, wmm enabled, power saving off, latest driver however no change. Is my wifi card broken?

The wifi card is connected to two antennas in a Samsung R70 notebook I wrongly assumed that the notebook already has three antennas, so I have to buy a third one. However I don't think two or three antennas should create such a big difference? Is there anything else I could try? The MAC layer speed for 802.11n is 120Mbps so everything below 50Mbps is just wrong.

thanks

 

[stevech]

35 megabits per second more than 11g can do, net of all IP and WiFi overhead.
11n in a 20MHz channel (non-channel-pair-bonding for 40MHz) will do about that speed too, in many conditions. I believe you said you're using iPerf for PC to PC testing, not via the Internet. It's easy to misconfigure iPerf.

11a/b/g/n data rate per frame (burst rate per frame) varies according to the signal strength and noise, and error rates. And it can be the weakest link in the bi-directional WiFi transfers. Most consumer routers don't list the average signal strength of the FROM client signal. Nor the competition for air time.

The WiFi air link bit rate achieved, for the signal condtions, is some number x. The net yield at the IP layer is typically 60% or so of the air link rate, due to WiFi overhead and duplexing and IP overhead, and if you use TCP there's another 10-15% of overhead.

Of course, your speed can be limited by many things, including competition for air time with neighbors' busy WiFi (if it IS busy), 2.4GHz non-WiFi devices, and the PCs' IP stack tuning.

 

[redcow]

thanks for your answer

Using a FTP file transfer to measure the speed next to the router, with the 2,4ghz network on 20mhz and the 5Ghz on 40mhz I get 4,2 MB/s ~ 33-35 Mbits.

the router shows the connected client speed as:

WALI_G 00:21:6A:C8:XX:XX 192.168.1.7 -57dBm -101 dBm

Switching the 2,4ghz network to 40mhz I improve to 9,1 MB/s (FTP) ~ 72-75Mbits,

WALI_G 00:21:6A:C8:XX:XX 192.168.1.7 -51 dBm -101 dBm (84%)
WALI_N 00:21:6A:C8:XX:XX 192.168.1.7 -57 dBm -118 dBm (47%)

However the N Network speed fluctuates between 21%-47%-85% signal strength, thats very strange.

If I'm connecting to the "G" network I get faster speeds than connecting to the N network (using channel bonding as both networks run on 40mhz).

All the tests were doing with no other client connected and no other wireless network in range. Switching both networks to 40mhz improved speed however its still too slow, 25 MB/s should be the maximum, and I get less than half speed sitting next to the router without any other clients and wireless networks? something is wrong here.

 

[thiggins]

First, get your client at least 10 feet away from the router or receiver overload will reduce throughput.

If you are getting 72 Mbps using 40 MHz mode, you're doing pretty good. There is NO N router I've seen that will give you 25 MB/s (where did you get that number?).

If I'm connecting to the "G" network I get faster speeds than connecting to the N network (using channel bonding as both networks run on 40mhz).

Huh? What "G" network. Maximum G speed will be in the mid-20 Mbps range. You said you can get 30+ Mbps on N, 20 MHz bandwidth.

Finally, don't get distracted watching signal strength or quality. Throughput is what counts. But realize that N has higher throughput variation than G due to higher bit error rates.

 

[redcow]

First, get your client at least 10 feet away from the router or receiver overload will reduce throughput.

Tried that too, didn't change much.

If you are getting 72 Mbps using 40 MHz mode, you're doing pretty good. There is NO N router I've seen that will give you 25 MB/s (where did you get that number?).

My fault I thought MAC layer speed would be 200Mbps however its only 120Mbps so the theoretical maximum speed for 802.11n would be 15MB/s, and for 802.11g 2,5 MB/s.
My 9 MB/s seem bad in confront to the possible max of 15MB/s taking into consideration that there are no other wireless network around or any other barrier (same room). I would expect at least 11-13 MB/s ?

Huh? What "G" network. Maximum G speed will be in the mid-20 Mbps range. You said you can get 30+ Mbps on N, 20 MHz bandwidth.

Sry for the confusion with "G" network I mean my 2,4 GHZ (with the ESSID "WALI_G") running on 40mhz instead of 20. Connecting to the N network ("WALI_N" obv running on 40mhz) my maximum is 33-35 Mbits ~ 4,2MB/s however if I connect to the 2,4ghz network ("WALI_G") running on 40mhz(!) I obtain 72-75mbits. Is that normal behavior? shouldn't I expect faster speeds using the 5GHZ network?

Finally, don't get distracted watching signal strength or quality. Throughput is what counts. But realize that N has higher throughput variation than G due to higher bit error rates.

Yeah I always measure with iperf + FTP file transfer.

With more distance (20feets and a wall between the outcome is even more disastrous, with my old Linksys WRT54GL (2,4GHZ) I obtained the maximum of 2,4-2,5MB/s so I thought with such a great outcome the 802.11n standard would improve it a lot. However now in the same position (Notebook) and even a better wifi card (now Intel Wifi link 5300 before Intel 3945ABG) using the N network (5GHZ essid "WALI_N" ) I get 11-13Mbits ~ 1,6MB/s and using the G Network (2,4ghz running on 40mhz) I get 52-60Mbits~ 6,9MB/s. Quite strange shouldnt the 5GHZ network be faster? Or is it because using the 2,4ghz network on 40mhz the wifi card uses channel bonding and achieves therefore much greater speeds?

So I'm trying to figure out if that are normal statistics using the WNDR3700 router with a decent wifi card, or if something is wrong in my setup (router, wifi card, some configurations settings...)

thanks for your time and help

 

[thiggins]

You can't take wireless link/PHY rates and infer maximum throughput from them. There is much more protocol overhead in wireless than wired.

See SmallNetBuilder's Wireless FAQ: The Essentials FAQ1 for typical maximum speeds.

5 GHz doesn't provide higher throughput than 2.4 GHz. But using 40 MHz bandwidth mode will provide higher speed than 20 MHz mode under strong to medium signal conditions. With weaker signal levels, speeds fall back to about the same with either mode. 40MHz bandwidth isn't not recommended in 2.4 GHz because it hogs almost the entire spectrum. 40 MHz bandwidth mode should be used in 2.4 GHz only if there are no nearby networks.

Bottom line, your speeds sound typical.

 

[redcow]

thanks it seems I was underestimating the protocol overhead and relying too much on the "300Mbps channel bonding dual radio hyper speed whatever" marketing garbage, if you say my speeds look good for a "realistic environment" I'm fine with it, thanks again for the clarification .

 

[stevech]

marketing garbage it is!
But it is helping them get a market saturated with 11g products to discard these and go 11n, only to find there's no substantial improvement in the internet experience for most users.