Bandwidth Test Results
Single Unit Performance
Although each system is packaged and sold to work in sets of three, we wanted to get an initial baseline of just how powerful an individual mesh access point could be. With Google and eero, this is easy, as all three units are the same. AmpliFi is different, of course, as it relies on a more powerful central base station with its less powerful MeshPoint satellites. So this test will compare the performance of the identical eero and Google units and the (on paper) more powerful AmpliFi HD.
The results show that even a single router (a mesh network minus the mesh, if you will) can provide adequate coverage in a small area, but things start to unsurprisingly break down at further distances. It's a close race between Google and AmpliFi, with only eero offering significantly lower performance at locations 3, 4, and 5.
While the results indicate that a single-unit setup wouldn't suffice in our test home, those with smaller homes or apartments could indeed receive adequate wireless coverage in such a configuration. eero and AmpliFi even advertise this possibility, noting that although you'd miss out on the whole "mesh" thing, single-unit owners could still take advantage of each platform's automatic configuration and maintenance. We wouldn't necessarily recommend that approach, as the cost of even a single mesh access point is still higher than most low- and mid-range traditional routers, but it's an option for those who'd like a future-proof system in the event that they move to a larger home or office, as additional mesh access points can be purchased later and added on to all three systems.
3-Unit Mesh Performance
Turning to an actual mesh setup, here's our initial look at how these systems perform with all three access points in action.
Right away we notice that something's wrong with Google. Both eero and AmpliFi turn in relatively good numbers at all locations, but the performance of Google Wi-Fi falls off a cliff at the more distant locations.
We performed the test with Google multiple times but the poor numbers remained. This led us to look at our testing process, which involved setting the MacBook down in the same position at each location, running the iPerf test, moving to the next location, and repeating. The problem with the Google test turned out to be that our MacBook remained connected to the initial access point in the office after moving to a new location, and so it was hanging on to that weaker signal instead of automatically shifting over to the stronger signal provided by the mesh extenders. The only way to correct this performance hit was to temporarily disable our MacBook's Wi-Fi chip, move to the new location, and then re-enable the Wi-Fi at which point it would pick up the closer, stronger network signal instead (we could also reboot the MacBook but this was a more time-consuming solution).
So we performed another round of tests using this disconnect/reconnect method for Google and the results were more in line with what we expected. Google's performance still lags behind eero and AmpliFi, but the gap is much more reasonable.
The funny thing is that eero and AmpliFi aren't doing anything special when it comes to handoffs between access points. There are standards like 802.11r which support wireless handoffs, but these aren't currently employed by eero or AmpliFi. Surprisingly, the handoff just seems to work for eero and AmpliFi while Google struggles.
The handoff issue won't be a dealbreaker for everyone, as many of our Wi-Fi devices such as smart TVs, wireless speakers, and PCs don't necessarily move around all that often. But the dream of mesh networking is that you can wander from room to room in your house or office and never lose signal on your smartphone, tablet, or laptop. Google currently has a major shortcoming in this regard, although it might be able to be addressed in a future firmware update.
As mentioned earlier, mesh networks have solved the wireless performance loss of previous systems by including dedicated radios for intra-mesh communication. Still, some homes and many offices already have a wired networking infrastructure, and both Google and eero allow you to hardwire one or more mesh points together (AmpliFi can't take advantage of this feature as its MeshPoints lack Ethernet ports).
Physically connecting your mesh access points, something called Wired (or Ethernet) Backhaul, can improve the performance of your network by both improving the intra-network communication speed and reliability between your mesh access points as well as freeing up the mesh point's second wireless radio to serve your wireless devices.
My new home is unfortunately devoid of wired Ethernet runs, so I instead picked up some really long Cat6 cables and just temporarily ran them down the stairs and through the rooms to each mesh point, a setup my wife just adored, I assure you.
The results below show the same locations tested again with the eero and Google units connected via gigabit Ethernet, with AmpliFi's earlier wireless-only results included for comparison.
As expected, we don't see much change in Locations 1 and 2, where the MacBook is still connected to the primary mesh point. At Locations 3, 4, and 5, however, there is a noticeable improvement as the now wired mesh points are free to combine multiple wireless antennas with a consistent, high speed wired connection. Although eero's performance in this configuration comes out on top, Google's system still sees impressive gains compared to its wireless configuration. Note, however, that we had to perform the aforementioned "disconnect/connect" procedure with Google to achieve these numbers.