Wireless Security Performance
Draft 11n routers are supposed to switch to supporting only "non HT" (High Throughput) modes, i.e. 54 Mbps maximum, when using WEP or WPA / TKIP wireless security. So I ran tests using WEP 128, WPA / TKIP and WPA2 / AES with the Intel 5300 client with 11n mode and Throughput Enhancement enabled, and with both settings disabled to put the adapter in 11g mode.
Unlike the Ralink-based Belkin N150, the WRT120N properly connected to the Intel adapter at a maximum 54 Mbps rate when it was acting as a draft 11n adapter and WEP or WPA / TKIP were set. But checking for throughput degradation was difficult, because during the tests, the link rate seemed to want to hang around 18 Mbps, even though I was in the same room (Location A), reducing the test througput.
However, I observed the same behavior (using a lower link rate with lower speed resulting) when I switched the Intel adapter to act as an 11g client. So the throughput reduction isn't because of any encryption overhead. It's simply due to some unoptimized interaction between the WRT120N and the Intel adpater.
Wireless Performance - Mixed Clients
One thing I neglected to do in my test of the Belkin N150 was to see how it handled a mix of draft 11n and 11g clients. Since Certified draft 11n routers don't do well with a mix of clients (throughput for both is severely reduced), I figured that routers using single-stream draft 11n would do no better. But this time, I thought it would be useful to see if a single-stream draft 11n router would do worse.
So I set up a test that connected my standard test notebook with the Intel 5300 AGN client in N mode and another notebook with an Intel 2915 ABG internal mini-PCI adapter to the 120N. The 11g adapter happened to be another Intel; I wasn't intentionally making this an Intel-only client test. The WRT120N was set to 20 MHz bandwidth, Mixed mode.
I then ran four tests:
- Downlink, G client starting first
- Downlink, N client starting first
- Uplink, G client starting first
- Uplink, N client starting first
If you open all four plots, you'll see that the draft 11n client has the lower average throughput in three of the four tests. That's not to say that the 11g client just sails along with nice steady throughput, as shown in the uplink plot in Figure 21.
Figure 21: Mixed client reduced performance example
Throughput sharing among mixed b/g and draft 11n clients tends to vary significantly depending on the exact mix of clients and router, not to mention surrounding networks. But it seems that, at least for this test case, the draft 11n client took more of a throughput hit than I have seen with other mixed client testing.