Testing performed by Tim Higgins
Trendnet initially provided a TEW-621PC V2.0R card for testing. But when initial testing revealed wide throughput variation which seemed to be caused by the card constantly changing its link rate, I started exploring. I found that the V2.0R card was using a Ralink driver, which didn't match the FCC board photo that showed an Atheros chipset. Long story short, Trendnet sent a V1.0R card, which is Wi-Fi Draft 2.0 certified and matches the FCC photo, so I started to test with that.
But, the Wi-Fi gods were still messing with me. Because when I went to connect the card to the Azimuth system, I found that the V1.0R card used different mini-connectors that I didn't have mating cables for. It's probably for the best, since the card had connectors mounted on only two of the three antennas!
So I fell back to using the Atheros-based D-Link DWA-652 Cardbus card that uses the same chipset as the Trendnet TEW-621PC V1.0R card. The card was inserted into a Fujitsu P7120 Lifebook (1.2 GHz Intel Pentium M, 504 MB) notebook running WinXP Pro SP2 with all the latest updates. I used the Atheros 6.03.107 driver and Windows Wireless Zero Config during testing.
I updated the router to 126.96.36.199 firmware and I left all factory default settings in place, except to set Channel 1 and disable the wireless auto-QOS WISH feature. Note that the 633GR properly defaulted to 20 MHz bandwidth operation upon power up.
Figure 9 shows a composite IxChariot plot of wireless uplink, downlink and simultaneous up and downlink tests with the router set to 20 MHz channel bandwidth mode. Testing was done with the router and notebook about 10 feet apart in open air sitting in my lab with no other networks in range.
The plot is hard to follow due to the high throughput variation. The two upper lines are the separate up and downlink tests, while the two bottom lines are the simultaneous up/downlink tests. Basically, throughput is a bit wonky in the default 20 MHz bandwidth mode, with best case throughput coming in at around 57 Mbps, in the downlink direction. But if you check the results for the DIR-655, you'll see similar results.
Figure 9: Up and downlink throughput - 20 MHz bandwidth
Figure 10 shows a composite of test runs with the Auto 20/40 MHz mode set, which produced significantly higher throughput. I monitored spectrum use during testing with the Cognio Spectrum Expert, which confirmed that Channels 1 and 5 were used when the tests were run.
Figure 10: Up and downlink throughput - 40 MHz bandwidth
These results are sigificantly better than the DIR-655's. In fact, they establish a new SmallNetBuilder high for maximum throughput from a draft 802.11n router with total up/down throughput of 117 Mbps and downlink-only throughput of 106 Mbps.
Since the 633GR is a virtual clone of the DIR-655, I didn't conduct extensive tests of security mode throughput. I only ran quickly through WEP 128, WPA/TKIP and WPA2/AES modes and saw throughput hits similar to those of the DIR-655.
I should note that although the 633R supports Wireless Protected Setup (WPS), the client application for its companion TEW-621PC card does not. So even if you buy the 633GR's "matching" card, you'll have to set up wireless security manually.
I ran no test of mixed STA or Legacy Neighbor performance, since the behavior of Draft 2.0 products has been well-established by now. There is no reason to expect that Trendnet (or more accurately, Atheros) has done anything to move the ball forward with the 633R.