Results: IxChariot, Same-Room, 5.0 GHz
Last up, we have Ixia’s IxChariot, which is probably the most widely used and trusted benchmark in wireless networking.
In our same-room 802.11ac testing for TCP traffic, Belkin sells the only router that falls flat. Even the next-lowest performer, Linksys, averages 160 Mb/s, which is stunning for TCP. Note that AirLive continues to impress at 189 Mb/s, solely on the basis of its beamforming. Just imagine when 11ac vendors cycle through their first wave of routers and decide to implement optional beamforming in their second-gen follow-ups in 2013 or 2014! Now, remember to approach those huge top-end numbers for Asus, Buffalo, and Netgear with caution. For example, here’s what was going on under the hood in Asus' IxChariot throughput graph:
See the plateau shift again? For sure, if we could count on Asus to average 320 Mb/s consistently, we’d bow down and worship this router. However, until we get a better fix on why these shifts keep happening, we’ll keep our adoration in check.
For Netgear, the situation is very similar. While its plateau averages are slightly higher than Asus', and its peak-to-trough range is similar, Netgear’s plateau shift arriving later in the test skews the average throughput downward.
We did not break out response numbers in these analyses since they were essentially inverse mirrors of our throughput charts. Take Linksys’s response graph (above) as an example. One can say that 0.5 second is the average response time, but this clearly depends on when in the test cycle we’re talking about.
If that’s not confusing enough, let’s switch to UDP traffic for the same-room set. Remember how we redlined most UDP results in PerformanceTest 7? IxChariot’s UDP script clearly limits bandwidth—so much so that our UDP numbers underperform our TCP results, which almost never happens.
Yes, our UDP throughput is roughly half of what we saw with TCP. However, regardless of the ways in which IxChariot constricts or relaxes data flow through its scripts, we still have an accurate relative ranking of routers all adhering to the same test process. And within these rules, we see all four of our serious 11ac routers averaging a statistical dead heat.
The plateaus we see in our TCP tests don’t appear here. On the other hand, we see a startling difference in throughput patterns. Consider these two graphs from Buffalo and Linksys. The averages are very close, but the patterns obviously differ.
Which graph is better? We would argue for Linksys. Whereas Buffalo seems to be struggling to keep pressed against a 118 Mb/s ceiling, Linksys has a very defined floor of around 112 Mb/s. When it comes to maintaining quality of service in a stream, the latter pattern is clearly preferable.