11g vs. 11g - Streaming
Atheros has said that Broadcom is harping on something common to all 802.11g products and that interference will also be generated by "normal" 802.11g WLANs operating in close proximity. So I first wanted to test to see if this were true.
Since Broadcom used a 2Mbps MPEG stream as part of their Comdex demonstration, I also wanted to include streaming as part of my testing. But I wanted a measure more quantitative than trying to describe the degree of jerkiness in a streaming video of The Matrix.
Fortunately, Chariot includes the ability to simulate video streams, and calculates the Lost Data in a stream. Due to the buffering and error recovery mechanisms built into streaming protocols, it's difficult to directly correlate a Lost Data number to the number and severity of glitches in a streaming video. But it's safe to say that lower Lost Data numbers are better than higher, and using Lost Data as a measurement is a hell of a lot easier than trying to describe how a video is breaking up!
Chariot doesn't include MPEG streams among its test scripts. So I had to make do using the NetMtgv.scr script, which emulates a NetMeeting 2.1 Video stream. I used UDP as the protocol and set the stream rate to 2.048Mbps.
Because the bit rate is limited to 2Mbps, the spectrum generated isn't very wide, which should limit the channel overlap, and therefore the amount of performance degradation.
My first round of tests were run with the following conditions:
Figure 9 shows the throughput plot for the test, which doesn't really tell you much. The blip near the end of the test appears in most tests and is caused by a periodic scan done as part of Broadcom's driver.
Updated December 6, 2003
Figure 9: Throughput for Atheros 11g vs Broadcom 11g - 2Mbps streams - 10ft