When you use jperf, take into consideration that the OS on both sides must be the same, i.e. windows xp or vista or 7 on both sides, OR Linux of the same distribution on both sides.

You have to take into consideration that the performance will never come as you like it to be, for example I have tested with Linksys WRT54GL v1.1, with Ubuntu desktop, and WPA2 PSK enabled, I got 13.3 Mbps for .11g. It was ~18 Mbps without WPA2 PSK. While the claimed performance is 54 Mbps.

I do not know a tool that is better than iperf (jperf is just a java interface for iperf) in regards to performance and feature set.

It has been approved by the World Bank, when we made a BW/jitter/delay/packet lose tests for the IIBN network in Iraq, it was the tool used, however, WB professionals approved the results.

PS: Usually Linux outperforms Windows.

I understand that you don't get anywhere close to the advertised data rates with wireless. Half-duplex plus associated 802.11 protocol overhead. 50% of data rate as actual throughput best case.

In this specific case with a typical noise floor and a lot of signal I was seeing things reported as what I believed to be abnormally low. Jperf (iperf) is what I've been using and adjusting the window size. Just wondered if there was another tweak I was missing that would improve things when I'm testing stuff.

I have reviewed the scrips that I used to test IIBN network, so what mainly comes in mind the following tweaks:

-- Increase the parallel streams to up to 3, you will not benefit of higher level

-- Test with both TCP and UDP, with changing the TCP window size and the Max segment size, as well as UDP packet size when use UDP

PS: I suggest to automate the process, I mean know the options from the command line, then create a shell script to automate the process, then automate the output.