I wasn't surprised to find that the mixed client performance of the 350N was virtually the same as that of the 600N. I set up my usual test using a second notebook with a Linksys WPC54G V3 11b/g card (Broadcom chipset). Both it and the Fujitsu notebook with WPC600N card were associated with the 600N, which was set to its 2.4 GHz Standard - 20 MHz Channel mode, to simulate what should be its proper "out of the box" settings. I then ran two IxChariot throughput.scr streams, alternating the STA that got on the air first.
Figure 17 shows how uplink throughput was shared when the 11n pair started first, while Figure 18 shows the 11g pair starting first.
Figure 17: Mixed 11n, 11g STAs - Uplink, N starts first
The results follow the pattern that I've seen with other draft 11n products,i.e. the 11g STA gets knocked down to 11b speeds. In this case the 11g throughput is knocked down to about 4 Mbps from its normal 20 Mbps or so, or about 80%. The 11n pair also gets a throughput hit, but still manages to run at about 40 Mbps when the 11g STA is on the air.
Figure 18: Mixed 11n, 11g STAs - Uplink, G starts first
The 11g client does a little better running downlink since it manages to run at around 10 Mbps when sharing the air with the WPC600N.
Legacy Neighboring WLAN Tests
I didn't run Legacy Neighbor / CCA tests, because I've already tested how Broadcom's Intensi-fi chipset reacts to a neighboring WLAN parked in its 40 MHz extension channel. Like every other Draft 2.0 11n product I've tested so far, I'm sure that the 350N would fail to fall back to using the legacy-friendly 20 MHz channel.
Starting with this review, I'm not going to test for CCA until the next 11n Draft revision, or when I review a product with a chipset that I haven't tested yet (which would be Ralink). I've yet to find a product that implements the fallback mechanism, and I doubt I will until the 11n Task Group decides whether they really want the mechanism to be mandatory.
As you've probably figured out by now, the 350N is essentially a single-band version of the WRT600N, with slightly lower wireless performance. But as I noted earlier, in real-world operation, the performance of the two siblings is close enough that you probably wouldn't notice much difference.
It comes down to how much you value having the option of operating in the less-crowded 5 GHz band. With current pricing, the dual-band 600N commands a $70 - $80 premium over the 350N. But as more dual-band products hit the streets early in 2008, that premium should shrink.