WiFi 7 is the kind of spec sheet that makes you want to believe. 320 MHz channels, 4096-QAM, Multi-Link Operation stitching bands together — the headline numbers promise a generational leap. I wanted to know how much of that survives contact with a real network and real walls, so I built one and measured everything.
The concept
We needed to follow the previous wifi 6's experiment environment. The idea was simple initially.
We thought at first it would be better to use the things that the original group used. But with time we found that what they did was not the best. We needed to expand out this project to make it more interesting. So we adapted, using kismet as the monitoring tool and using smb networking.
The build
We set both servers up with linux, wifi 7 cards and tested with a tp-link router. Which later we found was wrong. We then tried a p2p network as per our clients instructions. Then about 6 months into the project the client changed his mind again and allowed the upgrade to the ASUS router which we could then test on properly. We used the earlier 2.5ghz and 5ghz data from the tp link router running an earlier version of wifi 7. Then tried the 6ghz frequency with the Asus router.
For home/business networking 802.11be settings that matter most in practice:
- Multi-Link Operation (MLO): letting a client use the 5 GHz and 6 GHz bands at once.
- 320 MHz channels in the 6 GHz band — double the widest WiFi 6E channel.
- 4096-QAM for denser modulation when the signal-to-noise ratio is good.
All of this works on the correct router, for us the earlier tp-link router wasn't correct, but on the ASUS router this would be acheivable.
What I measured
- Peak throughput close to the AP was dramatically higher — the wide channels do exactly what they say when conditions are clean. 6ghz fastest close range only, 5ghz med range (still faster than wifi 6), 2.5ghz longer range(faster than wifi 6). The main improvements to wifi 6 are really the protocol.
- Latency under load was the real win. MLO's ability to dodge a congested band kept things responsive.
- Range-limited spots narrowed the gap fast — wide channels need clean spectrum and good SNR.
So — was it worth it?
For raw "my speed test is bigger" bragging rights? Only near the AP. For a network that stays responsive when everything is happening at once? Genuinely yes — MLO is the feature I'd actually miss. But the biggest gain was the process: designing the topology, reasoning about RF, and documenting every step so the results are repeatable.