The complaint at the lake house in Muskoka was the universal one: “why does the internet suck?”
It shouldn’t have. The fiber line there is 2.3 gigabits, symmetric, and I verified it — the gateway’s own wired speed test clocked 2.31 down and 2.37 up. The internet was innocent. But every single WiFi device in the place topped out at almost exactly 90 megabits. Different phones, different laptops, different speed test servers, both WiFi bands: 87, 89, 92, 90. Meanwhile a wired test run at the same moment pulled 2,100.
Everything lands on 90
That consistency is the tell. Random interference gives you random numbers. When everything lands on the same ceiling, something in the path is that ceiling — and ~90 megabits of real throughput is the exact signature of a 100-megabit Ethernet link hiding somewhere it shouldn’t be.
One badly punched jack
Found it in the access point list: the in-wall AP had negotiated its uplink at 100 Mbps instead of a gigabit. Gigabit Ethernet needs all four wire pairs in the cable; when a link comes up at 100, it usually means only two pairs are actually working — which is the classic signature of a badly punched keystone jack at one end of the run. Whoever terminated that wall jack got two pairs seated and called it a day.
Twelve devices, one straw
And here’s the multiplier: that particular AP was carrying 12 of the roughly 15 wireless devices in the house. The other AP — with a healthy gigabit uplink — had three. So most of the household was sharing one 100-megabit straw while a 2.3-gigabit pipe sat behind it, bored.
The fix is not software. It’s a punch-down tool and five minutes of re-terminating both ends of that run, all eight conductors this time.
The smart-queue detour
I’ll also confess the detour, because it was a proper rabbit hole: before finding the jack, I spent a while treating this as a congestion problem and applied a smart-queue traffic shaper. In testing it looked heroic — latency under load dropped from 2.8 seconds to 284 milliseconds. It also quietly throttled the actual 2.4-gigabit upload to 75 megabits, which is like fixing a squeaky door by welding it shut. Fully reverted, along with an hour I lost to a flaky speed-test endpoint that reported a phantom 87-megabit upload and sent me hunting a line problem that didn’t exist.
Two rules survived
Two rules survived the trip. Judge an internet connection only from the gateway’s own wired test, never from a phone on WiFi — the WiFi is almost always the thing you’re actually measuring. And when every device agrees on a suspiciously round number, stop testing and start looking for the link that is that number.
The internet was fine. This time, the wall jack was the thing I should have interrogated first.