Wireless Performance
Test and analysis by Tim Higgins
Testing the Extreme's wireless performance presented some unique challenges. While Ixia has a MacOS endpoint for IxChariot, it is PowerPC and not Intel-based. So I chose to install BootCamp on the top-of-the-line black MacBook that Apple sent along with the Extreme and loaded a copy of XP Pro SP2. (I actually ended up running some quick throughput tests using the PowerPC IxChariot endpoint, which seemed to produce results similar to those produced by XP.)
The router already had the latest 7.2.1 firmware and I left all factory default settings in place, except as noted.
Maximum Throughput - 2.4 GHz
To start, I ran some close range open-air IxChariot tests to look at maximum performance and throughput variation. This also provides baselines to check the Azimuth results against. As with the routing section tests, I ran all open-air wireless tests with Win XP SP2 running via BootCamp in the MacBook.
Testing was done with the router and MacBook about 10 feet apart in open air sitting in my lab with no other networks in range.
Figure 22 shows simultaneous up and downlink results in the default 2.4 GHz mode (b/g compatible), which come in at around 86 Mbps of total throughput. Running up and downlink separately yielded results of 78 and 66 Mbps respectively.
Figure 22: Up and downlink throughput - 2.4 GHz band, 20 MHz bandwidth
Since Apple has chosen to lock out 40 MHz bandwidth (channel bonded) operation in the 2.4 GHz band, I couldn't test that mode. But I did check the "n only" mode and saw little performance difference.
Maximum Throughput - 5 GHz
I next checked throughput in the 5 GHz band. In this band, Apple defaults to using the 40 MHz bandwidth and "11a compatible" modes, so that's what you see in Figure 23. The total throughput of 124 Mbps is pretty impressive and better than the 103 Mbps of total up/down throughput turned in by the only other dual-band draft 11n router now available, the Buffalo WZR-AG300NH. I should note, however, that this superior performance was not reproduced in the Azimuth-based throughput vs. path loss testing. More on this shortly.