"Slow Wi-Fi" After a Fiber Upgrade
An office reports "slow Wi-Fi" after a 1 Gbps fibre upgrade. The user's laptop shows a 144 Mbps negotiated rate, and the vendor brochure promises "up to 1.2 Gbps." An internet speed test returns 90 Mbps down. iperf3 to an on-premises server over the same AP and same channel returns 210 Mbps. The serving AP shows 55% channel utilization, 12% retries, and 38 associated clients, of which 18 are on 2.4 GHz. The AP uplink negotiates 1 Gbps full duplex; the PoE injector is 802.3af (15.4 W). The customer demands a truck roll to replace the access point.
1. What does the pair of tests establish, and what does it rule out?
Read the worked answer
It rules out the AP and the RF link as the binding constraint on the 90 Mbps result. The same client, the same AP, the same channel and the same airtime produced 210 Mbps to an internal server. If the AP were the bottleneck, both numbers would be low and roughly equal. A 120 Mbps gap between what the AP can deliver and what the internet test returns has to sit downstream: router, firewall, traffic shaper, WAN handoff, DNS, or the speed-test server and route.
The 90 Mbps figure is diagnostic, not just disappointing. It sits almost exactly at 100BASE-T line rate, which makes a 100 Mbps link or a 100 Mbps policer the leading hypothesis - a switch port, the PoE injector's data path, the router's WAN interface, or a shaper left at the old circuit's rate.
Therefore: a truck roll to replace a functioning AP is not supported by the evidence, and would likely change nothing while resetting the customer's expectations to "the new one will be faster."
2. The customer points at "up to 1.2 Gbps" and at the laptop's 144 Mbps. Explain the gap without hand-waving.
Read the worked answer
The 144 Mbps figure identifies the configuration. That is the top rate for a 2×2 client on a 20 MHz 2.4 GHz channel. The laptop is on 2.4 GHz, 20 MHz - a configuration whose ceiling is around 144 Mbps PHY and roughly 90–100 Mbps of goodput. Note that this is close to the observed speed-test result, which is another reason not to accept the speed test as a clean measurement of the internet path.
The 1.2 Gbps figure describes a different product. It assumes a 5 GHz wide channel (80 or 160 MHz), multiple spatial streams and short guard interval with a capable client. None of that is in play here: the client is 2×2 and on 2.4 GHz.
So both numbers are consistent with the same ordinary connection, and neither is evidence of 1.2 Gbps capability in this deployment. Separately, 18 of 38 clients on 2.4 GHz at 55% utilization explains why 2.4 GHz clients feel slow - but that is a distinct issue from the 90 Mbps internet complaint, and conflating the two is how this becomes a hardware argument.
3. What are your next three checks before anyone touches hardware, and what result would make you reverse your call?
Read the worked answer
Checks, in order:
- End-to-end link negotiation - switch port to PoE injector to AP to router WAN interface. Look for a 100 Mbps segment, a duplex mismatch, or an injector that only passes 100 Mbps.
- Shaping and policy - any policer, QoS or bandwidth profile still configured at the pre-upgrade rate on the firewall or router.
- Wired bypass - run the same speed test from a wired port on the same VLAN, plus a DNS resolution timing check, to remove Wi-Fi from the path entirely.
Also worth one glance, but not first: the 802.3af injector delivers 15.4 W at the PSE. Confirm the AP has not disabled a radio or throttled features for power budget. That is a real constraint, but chasing it before the wired test inverts the evidence order.
Reverse the call if: the wired test also returns ~90 Mbps (confirming the upstream path and vindicating the customer's symptom, though not their diagnosis), or if wired returns ~900 Mbps and iperf over Wi-Fi collapses under concurrent load from the 38 associated clients - in which case contention is genuinely in play and the 210 Mbps single-client figure was an unrepresentative best case.