• "Signal Quality" readings were not available from the WinXP Zero Configuration client used
• Testing was done with a DrayTek Vigor540 Wireless 11g CardBus Adapter in a WinXP Home Dell Inspiron 4100 laptop unless otherwise noted
The throughput test results in Figure 18 show best-case performance of 24Mbps, which is comparable to other 802.11g products running without any of the myriad throughput enhancement technologies.
Note that the lowest result was obtained in Location 3, which is one floor below and at the opposite end of the test building.
Though a check of the Vigor540 test partner client card's Advanced Network properties showed a Nitro mode (Conexant's throughput-enhancement technology) entry that was set to 1, I suspect that neither Nitro nor Nitro XM is enabled in the 2900G itself. If it had been, I would have expected to see more like 27Mbps under the same test conditions.
Location 1 (best case signal) runs with WEP128 then WPA-PSK (TKIP) enabled yielded average 1 minute speeds of 23.0 and 18.1Mbps respectively. These translate to throughput reductions of about 5% and 25% respectively, when compared to the "naked" best-case 24.1Mbps result.
The 25% reduction for WPA (TKIP) is right in line with tests I've run on other PRISM-based products and appears to be baked into the chipset. Unfortunately, the PRISM chipset doesn't support the alternative AES encryption that Broadcom provides as a way around the problem in their chipset. So if you want to use WPA with the 2900G, you'll unfortunately pay a high throughput penalty.
802.11g Wireless Performance Test Results
|Test Description||Signal Quality (%)||Transfer Rate (Mbps)||Response Time (msec)||UDP stream|
|Throughput (kbps)||Lost data (%)|
|Client to AP - Condition 1||0||24.1
|Client to AP - Condition 2||0||12.9||1 (avg)
|Client to AP - Condition 3||0||8||2 (avg)
|Client to AP - Condition 4||0||13.6||1 (avg)