I used Ixia's IxChariot and the standard throughput.scr script to generate TCP/IP traffic for testing. The only change made to IxChariot script defaults was to change from 100,000 Byte to 1,000,000 Byte test size. Each test was run for one minute.
I started by plugging each set of kit adapters into an outlet strip to test for best-case throughput. (Yes, it's ok that I'm using an outlet strip because I'm only going outlet to outlet. The main reason why powerline product installation instructions tell you to not use outlet strips is that many have line filtering that can reduce your performance.)
Figure 25 shows a summary of best-case transmit performance. All the adapters except the Linksys PLK300 kit run in the mid-to-high 40 Mbps range. The Linksys, which happens to also be the only product to use an older, second generation chipset, only reaches 37 Mbps.
Figure 25: Best case transmit throughput - five HomePlug AV adapters
Figure 26 shows receive test results, which trend about 5 Mbps lower. Note that both receive and transmit plots show large throughput dropouts, which could affect HD streaming, depending on how good a buffer your playback mechanism has. Average throughput, however, should be more than enough for even 1080p playback.
Figure 26: Best case receive throughput - five HomePlug AV adapters
I've found with MoCA and some wireless products, that running multiple simultaneous streams can reveal higher aggregate throughput. So I ran four transmit streams using the NETGEAR XAVB2501 kit, with the results shown in Figure 27.
Figure 27: Four simultaneous transmit streams - NETGEAR XAVB2501
I was able to go from 48 Mbps using a single stream to just shy of 80 Mbps using four streams—about a 66% gain. A bit more experimentation showed that two streams got me up to 73 Mbps and three achieved the same 80 Mbps as four streams.
Note this multiple stream performance isn't unique to the NETGEAR adapters, since it's part of how HomePlug AV works.