All performance tests were done with 22.214.171.124.382_11014-g32b9d2b firmware and 126.96.36.199.28 Android app using select tests from the Version 10 Router process.
The simple iperf3 WAN to LAN and LAN to WAN throughput tests came in at a surprisingly low 523 and 529 Mbps, respectively. I suspected AiProtection and Adaptive QoS could be causing lower throughput, so I disabled both and reran the tests. This raised WAN to LAN throughput up to 638 Mbps and LAN to WAN to 614 Mbps. So if you have a gigabit internet connection, you'll need to give Lyra a pass, at least until ASUS implements AP mode and you can use another router.
Routing throughput - iperf3 method
The HTTP test method is better at showing performance differences under load, downloading four different size image files with 2K concurrent connections. The plot below compares Lyra with eero Gen 2, Orbi "mini" and Google Wi-Fi. Lyra doesn't do well in any of the test cases. Shutting off AiProtection and Adaptive QoS improved things considerably, yielding scores of 12%, 24%, 65% and 88% for 2KB, 10KB, 108KB and 759KB file sizes, respectively.
Routing throughput - HTTP Score comparison - WAN to LAN
Plot key file size: [A] 2 KB, [B] 10 KB, [C] 108 KB and [D] 759 KB file
Results are very similar for uplink. Disabling AiProtection and Adaptive QoS raised scores to 13%, 30%, 69% and 91%.
Routing throughput - HTTP Score comparison - LAN to WAN
Lyra was run through our new Wi-Fi System test process, letting Lyra use whichever channels and bandwidth it chose. Here's what Lyra looked like in the octoBox 18" test chamber. Because Lyra's three-radio design is most similar to eero Gen 2 and its pricing is closest, I'm going to focus comparisons to these two products.
ASUS Lyra in test chamber
Throughput vs. Attenuation (RvR)
The Rate vs. Range or RvR benchmarks look at how throughput varies with decreasing signal. This test is done on the root node, so is a best-case view and does not include any effects from backhaul links.
The 2.4 GHz downlink plot shows eero and Lyra tracking pretty closely. Both ran with 40 MHz bandwidth and reached full 400 Mbps maximum link rate at the beginning of the plot.
Throughput vs. Attenuation - 2.4 GHz downlink
2.4 GHz uplink, again shows similar performance for both products, except for a dip in eero's throughput at the stronger signal levels at the start of the plot.
Throughput vs. Attenuation - 2.4 GHz uplink
The two products diverge on the 5 GHz downlink benchmark, with Lyra showing a much faster throughput decline than eero.
Throughput vs. Attenuation - 5 GHz downlink
5 GHz uplink shows a similar result. These results would seem to indicate eero would have better 5 GHz range than Lyra. Of course, since both operate in a multi-node mesh, the difference might not be immediately obvious. In fact, just as in multi-AP systems, long range in mesh systems isn't necessarily an advantage. If nodes overlap too much, devices will have a difficult time figuring out when to roam and where to roam to.
Throughput vs. Attenuation - 5 GHz uplink
On the other hand, since mesh systems lean heavily on 5 GHz for backhaul, you'd want as much 5 GHz range as you can get. It's not an easy tradeoff for system designers.