Wi-Fi Slow at Night? Check Port Saturation and Congestion
Wi-Fi slow at night? Compare idle and loaded tests, check port saturation and link rates, and separate household queues from congestion or weak room coverage.
Wi-Fi slow at night can involve port saturation: traffic approaching the capacity of an Ethernet port, WAN connection or shared uplink, with packets waiting in queues. Here, port means a network interface, not a TCP or UDP service port. Evening household demand, neighbor airtime, a slow physical link, provider congestion and weak room coverage need different remedies; compare the same device at quiet and busy hours before choosing one.
Separate port saturation from a slow negotiated link#
A port that negotiates at 100 Mbps instead of an expected 1,000 Mbps creates a fixed ceiling. It can limit a wired test in the morning as well as at night. Evening traffic may make that limit more noticeable, but the link rate itself is different from demand saturating it. Check the connected device's Ethernet status and the router's relevant LAN, WAN or uplink status rather than assuming every socket is gigabit.
Follow the path through laptop, cable, switches, router and modem or ONT. A faster laptop port cannot compensate for a slower link elsewhere. For an unexpected 100 Mbps link, check specifications and try a known-good cable or supported port. Do not force negotiation settings without manufacturer instructions.
Four measurements describe different things:
| Measurement | What it describes | What it does not establish |
|---|---|---|
| Ethernet link rate | The negotiated physical link, such as 100 or 1,000 Mbps | Your internet service capacity or current traffic |
| Internet service capacity | The provider's download and upload limits | A cable's negotiated speed |
| Wi-Fi PHY rate | A radio connection's signaling rate | Application throughput or available airtime |
| RSSI | Received radio signal strength | Queue delay or port utilization |
For example, a 20 Mbps upload service can fill while its router WAN Ethernet link remains at 1,000 Mbps. The service limit is a separate bottleneck; a gigabit label does not supply a gigabit upstream allowance. Likewise, a strong Wi-Fi signal does not make a loaded WAN queue disappear. If every hour shows the same wired ceiling, start with the gigabit speed chain before blaming the evening schedule.
Run the same-device idle, load and pause comparison#
At a quiet time and the usual bad hour, use the same laptop, test server and positions. Record connection type, Wi-Fi band, Ethernet rate, download, upload and ping. Changing the server also changes the route being tested.
- Wired baseline: connect to the router, turn laptop Wi-Fi off and record idle ping without deliberate test load. Note household streams, backups or updates still running.
- Wired under load: run a download test, watch ping, then test upload separately. Record which direction makes the delay rise. A speed test creates traffic; it is not an idle reading.
- Household pause: briefly pause optional backups, game downloads, photo sync and streams at the bad hour, then repeat wired measurements. Keep essential work and safety devices operating; note traffic you could not pause.
- Nearby Wi-Fi: disconnect Ethernet and repeat beside the router, using the same band where possible. This checks the radio path without the far room's walls.
- Far-room Wi-Fi: repeat at the working seat and compare with the nearby reading. Test the actual call, game or transfer, not only downloads.
The patterns narrow the investigation; they do not automatically identify one failed component.
| Repeatable pattern | What to investigate | Next check |
|---|---|---|
| Wired improves after optional household traffic stops | Household demand or loaded queues | Compare upload and download load separately |
| Wired remains worse at night after the pause | Provider path or router/modem limits | Repeat on several nights and review the wired path |
| Wired holds; nearby Wi-Fi falls | Airtime, interference or router radio behavior | Compare band, channel and router load |
| Nearby Wi-Fi holds; only the far room struggles | Coverage path and signal margin | Test the same seat nearer a doorway |
| Wired has a similar ceiling at both hours | Link rate, service limit or device capacity | Check each link and the service tier |
Estimate utilization from one direction's byte counter#
If your router exposes byte counters, identify the interface and direction. WAN receive/transmit and laptop receive/transmit describe different boundaries; do not mix their labels. Record two readings and elapsed time. A graph without counters is insufficient for this calculation.
This is an illustrative calculation, not a measurement from a real home. Suppose one interface's receive counter increases by 112,500,000 bytes over 10 seconds:
112,500,000 bytes / 10 seconds × 8 = 90,000,000 bits per second.- In decimal networking units, that is 90 Mbps.
- Against a 100 Mbps link:
90 / 100 × 100 = 90%average receive utilization. - Against a 1,000 Mbps link:
90 / 1,000 × 100 = 9%average receive utilization.
Calculate transmit separately using its own counter change. Do not add receive and transmit percentages to claim one full-duplex direction is overloaded. These are simplified estimates; protocol overhead is not fully represented. Cisco's interface-utilization explanation describes counter differences and separate input/output calculations.
The example does not turn a 20 Mbps upload service into a 90 Mbps upload connection. A LAN or downstream counter and the service's upstream capacity refer to different boundaries. Choose the capacity for the boundary and direction being evaluated.
If the counter resets, wraps or changes meaning during the interval, discard this simple calculation and collect a valid new interval. A long average can hide short bursts. Neither one counter sample nor a speed test proves which port is congested: check repeated timing, queue delay, packet errors, cable condition and router load before assigning the cause.
Match the remedy to the bottleneck#
A throughput number and responsive calls are different outcomes. As an illustrative example, idle ping might be 15 ms and loaded ping 200 ms. That large change suggests investigating delay under load, not a universal cutoff or proof of one queue's location. Record the destination and traffic direction. Cloudflare's explanation of bandwidth and working latency describes why traffic waiting in buffers can delay interactive applications despite an acceptable throughput figure. Gaming lag and bufferbloat covers the practical follow-up.
Schedule backups, photo sync and updates during quiet hours while preserving essential calls and uploads. Streaming, calls and backups use different amounts of capacity; check each application's requirements. A backup can fill the smaller upload allowance while download capacity remains available.
Where supported, consider QoS or SQM. Follow your router's documentation for capacity settings and configure download and upload separately. Bursty traffic, a variable service rate and the router's processing capability can require different values; there is no universal setting. Retest loaded latency and usable throughput after each adjustment; enabling a feature is not proof it helped.
Wire a fixed high-demand device where practical. Moving a television or desktop to Ethernet reduces its Wi-Fi airtime use, but the same internet download still contributes to WAN demand. Do not promise that wiring one device cures a saturated internet connection.
If nearby Wi-Fi worsens while wired performance holds, inspect radio conditions. On 2.4 GHz, channels 1, 6 or 11 at 20 MHz are common options where local channel rules permit them. A quieter 5 GHz channel or narrower width may help in a dense building. Follow the channel-congestion guide and compare before and after at the bad hour. A visible neighboring network is not evidence that it caused your slowdown.
When only the far room fails, investigate walls and placement. An extender can help a weak coverage path if it still has a usable connection to the router; it cannot add provider capacity or remove competing airtime. Use Wi-Fi slow in one room to verify that pattern before adding hardware.
Keep an evening log before buying equipment#
Repeat on two or three typical weekday evenings. Log room, time, server, band, Ethernet rate, idle/loaded ping, both traffic directions and active jobs. This separates an effective remedy from a coincidentally quieter night. Recheck next week if the pattern changes.
Change one thing at a time, then leave it stable long enough to compare several evenings. Record optional jobs you paused and restore normal schedules afterward. An immediate improvement after a restart or setting change is useful to note, but it does not isolate the cause. Watch for packet errors or router load alongside latency spikes; those alternatives matter even when the average byte rate looks modest.
If wired performance remains poor with household demand paused and the local path checked, contact the provider with dates and results. Ask about outages, utilization and service limits rather than alleging throttling. Router/modem limits remain possible; the ISP test helps organize the evidence.
Plan the coverage change in Range Up#
For coverage, download Range Up on iPhone or Android. Free tools let you draw the floor plan and walls, place the router, mark a target and view simulated coverage. Premium adds extender placement, a recommended spot and estimated Before / After results. The app does not measure live RSSI, read port counters, run speed tests or configure routers. Use the Wi-Fi extender placement guide, then validate the plan with real evening tests.
Frequently asked questions#
Why is Wi-Fi slow at night but fine in the morning?
Evening demand can load household connections, provider capacity or shared Wi-Fi airtime. Compare wired, nearby and far-room Wi-Fi at both hours, then pause optional household traffic. A decline alone does not identify the provider.
What does port saturation mean for Wi-Fi?
Traffic approaches an Ethernet, WAN or shared uplink capacity, potentially creating queues. This concerns network interfaces, not TCP or UDP service ports. Strong Wi-Fi signal does not remove that bottleneck.
Is a 100 Mbps Ethernet link the same as saturation?
No. A 100 Mbps link is a physical ceiling; saturation depends on traffic using it. The ceiling can exist all day and become more noticeable with evening demand.
Does turning off other devices make a difference?
Pausing uploads, backups, downloads or streams reduces demand. Improved wired performance indicates household load contributes. Wiring devices instead reduces radio airtime but does not remove their internet demand.
Will a faster internet plan fix slow Wi-Fi at night?
It may help when service capacity is the limit. It will not repair a slow cable link, crowded airtime or weak room coverage. Compare those paths before changing plans.
Will an extender help with slow Wi-Fi at night?
It can help a weak room path when the extender has a good router connection. It does not resolve port saturation, WAN queues or provider contention. Verify it under real evening conditions.