2.4 GHz STA Only
I was pretty frustrated with AiMesh at this point. Its insistence on using the same channels on both radios on all three nodes, combined with its congestion-maximizing use of the same band for STAs and backhaul, kept me from getting anything like the throughput I would expect from a three node system.
Forcing the STAs to the 2.4 GHz band was my last hope for using a three-node AiMesh system with Wi-Fi backhaul. With the STAs all on 2.4 GHz, they should stop battling with their own backhaul.
Mean application latency comparison - all STAs forced to 2.4 GHz
But the last hopes for a workable three-node AiMesh system with Wi-Fi backhaul were dashed here. Although forcing all STAs to 2.4 GHz (3 nodes WiFi bkhaul No bndstr 2.4 GHz in the chart) did keep them from congesting with their own backhaul, the greater range and penetration of 2.4 GHz meant they maximally competed for bandwidth with each other instead.
All four STAs had higher latency when forced to connect to 2.4 GHz radios, but STA A and B were significantly worse. STA A, B and D all connected to the upstairs living room RT-AC68U. Only STA C, sitting right next to the same node, connected to the router. To make things worse, STA B changed to the downstairs RT-AC68U during the 16 Mbps rate test, moving from a one to two-hop connection.
Note the 5 GHz connection (3 nodes WiFi bkhl No bndstr 5 GHz in the chart) is the same data shown in the previous 5 GHz STA chart. This produced the worst latency of 5.3 seconds in this comparison.
I really like the overall premise of AiMesh - the idea that you should be able to add more routers to the router you already have and like, instead of needing to chuck it out and invest in a completely different mesh kit.
If you've currently got one AiMesh-capable ASUS router that you're really happy with, adding a second in AiMesh mode instead of just using any old Wi-Fi extender isn't the worst idea. The comparison below brings together the results from configurations that provided lower latency for all STAs than the the solo RT-AC1900P.
If you don't have the option of Ethernet backhaul, you'd best stick to only adding only one AiMesh node (2 nodes WiFi bkhl). The results show lower mean application latency for all STAs with worst case latency of a bit over 850 ms for the most remote downstairs bedroom STA B.
Mean application latency comparison - best cases
If you have the option of Ethernet, my tests (3 nodes Eth bkhl No bndstr 5 GHz) show you can move up to two additional AiMesh nodes, but you'd best force your devices to all connect on 5 GHz. Doing all this and maybe with a little bit of good luck, I was able to keep all latencies for all STAs below 400 ms.
The bottom line is that the data show that AiMesh, at least in its current form, isn't up to managing STAs and backhaul connections when both use the same pair of radios in each mesh node. And worse, AiMesh doesn't use Ethernet backhaul effectively. So even if you don't need the (Wi-Fi) mesh part of AiMesh and use Ethernet to connect all its nodes, it will give you worse performance than manually converting the routers to APs and configuring them yourself.
There are still some experiments I could have tried. For example, I could've tried dropping the 2.4 GHz transmit power to its absolute minimum and running the 2.4 GHz-only tests again. AiMesh is also clearly not really up to intelligently managing multiple-link backhaul; so I could also have tried pulling the downstairs RT-AC68U upstairs, and setting it up somewhere with a direct link to the RT-AC1900P router instead. I also could have introduced a tri-band router (or two) into the mix, to see if AiMesh could effectively use a second 5 GHz radio for dedicated backhaul.
But I didn't do any of those things because the tests I did run, which have provided the most detailed examination of AiMesh performance you'll find anywhere, show there are much better, less frustrating and less expensive options for Wi-Fi mesh systems.
Testing I've done for another publication has shown Orbi, eero, and Plume "Superpods" were able to handle the same four STAs using Wi-Fi backhaul (which is why you buy a mesh Wi-Fi system in the first place) and produce comparable mean application latencies. While these compeing system's ain't cheap, neither is the AiMesh configuration I tested. But they have the advantage over AiMesh of not requiring Ethernet connection to provide as good or better results.
|- RT-AC1900P (x1)
|- RT-AC68U (x2)||$280|
|NETGEAR Orbi||RBK53 One router, two satellites||$520|
|eero||eero + two Beacons||$398|
Table 1: Wi-Fi mesh system pricing (click links for latest prices)
And if your spouse is a believer that good Wi-Fi should not clash with the home decor, then AiMesh is likely to lose that battle.
Which do you think will have higher SAF?
So again, If you don't already own an ASUS router and you're in the market for a multiple access point system, there are frankly better options. If you have or plan to deploy wired Ethernet backhaul to your access points, purpose-built wired access points such as Ubiquiti UAP-AC-Lite, TP-Link EAP-225v3 or other APs reviewed in the 2x2 AC Access Point Roundup series are both less expensive and higher performance than AiMesh routers.
If you can't or don't want to use Ethernet backhaul and just want mesh, mesh Wi-Fi systems such as eero with a pair of Beacons still cost at least a little less than most AiMesh options, and tend to be more stable, easier to live with, and higher in Spouse Acceptance Factor.