The practical guide

Wi-Fi or internet connection? A controlled Ethernet comparison

Published by InternetSpeedGeo · Reviewed

1. Begin with the problem you actually experience

×

Write one sentence describing the problem before running a test. For example: video calls freeze in the upstairs office, but web pages load normally near the router. This is more useful than saying the internet is bad. Include where the problem occurs, which device is affected, and whether it happens continuously or only at certain times. If a particular service is affected, record that too. A browser measurement and a video meeting travel to different destinations; one can work well while the other does not.

Decide what would count as improvement. For a call, fewer interruptions may matter more than a larger download number. For sending a large file, upload speed and uninterrupted completion matter. For a room with unreliable coverage, maintaining a usable connection may be the first goal. Keep these observations alongside numerical results so that the investigation remains about usability, not winning a speed contest.

Start with equipment already available. You do not need to purchase a router, a subscription, or an adapter to discover useful patterns. If Ethernet is unavailable, the near-versus-far comparison still helps, but label the wired baseline as unavailable. Do not describe a comparison between a laptop on cable and a different phone on Wi-Fi as a controlled test of wireless performance alone.

2. Understand the path you are comparing

×

Your test involves several links: the browser and device, the local connection to your router, the router's internet connection, networks beyond your home, and the measurement endpoint. Switching from Wi-Fi to Ethernet changes the local connection, but it does not remove the router, provider, or distant endpoint. A better Ethernet result therefore supports investigating the wireless portion. It does not identify a specific interfering device or prove that every other part of the connection is perfect.

Likewise, equally slow wired and wireless results do not immediately establish a provider fault. Both tests may share a busy router, a limited adapter, background traffic, a VPN route, or an overloaded device. Think of the comparison as narrowing possibilities rather than issuing a verdict. Each next check should separate two explanations that are still plausible.

InternetSpeedGeo uses Cloudflare's measurement library for its main browser test. Consult this site's methodology for the configured measurements rather than assuming that every capability in the upstream library is enabled here. A local Ethernet link speed shown by your operating system and a browser internet result describe different things. The former is a link characteristic; the latter is a measurement across the longer path. Keep those labels separate in your notes.

3. Prepare a fair baseline without disrupting the household

×

Choose a short period when you are allowed to test. Tell anyone sharing the connection that the measurement transfers real data and may briefly compete with their work. Do not run it during an important call, a remote examination, or a critical upload. Pause your own backups, game downloads, cloud synchronization, and streaming where practical, but do not disconnect someone else's equipment without permission. Record any activity you cannot pause instead of pretending the connection was idle.

Keep the same device, browser, test page, and power conditions throughout the initial comparison. Close unnecessary demanding applications. Wait for the current test to finish before starting another. Keep the browser visible, avoid moving the device during a measurement, and do not switch networks halfway through a run. These simple habits make the observations easier to compare without requiring specialist equipment.

Record the date, approximate time, location, connection type, and VPN state before each run. Give locations descriptive labels such as office desk and living room beside router. If you know the wireless band or mesh node, include it; otherwise write unknown. Unknown is a valid observation. A network name containing a number is not reliable proof of the actual band or access point in use.

4. Read the units before interpreting a difference

×

Download and upload are shown in megabits per second, commonly written Mbps or Mb/s. File-transfer applications may instead display megabytes per second, written MB/s. There are eight bits in a byte: 80 Mbps corresponds to 10 MB/s before accounting for overhead and changing conditions. A file application showing 9 MB/s is not necessarily performing nine times worse than a speed test showing 80 Mbps. First put the numbers in the same unit.

Ping is a delay measurement in milliseconds, not another speed in megabits. A connection can move a large amount of data while still responding slowly to individual interactions. Jitter describes variation in latency, and its calculation depends on the tool. Do not compare differently defined jitter values as if they were produced by identical methods. Cloudflare's latency reference explains the distinction between delay, bandwidth, and achieved throughput.

For this investigation, compare download with download, upload with upload, and the same latency measurement with itself. Do not combine them into an unexplained average. Record an unsuccessful test as failed or incomplete, not zero Mbps. A zero inserted for a missing observation can distort a median and make a connection appear worse than the evidence supports. Keep a screenshot or exported result when available.

5. Run the three-location experiment

×

First measure at the place where you normally experience the problem. Keep the device in its usual position rather than holding it against the ceiling to obtain a better number. Then move the same device near the router, in an open position with a clear path, and repeat. Finally, if available, connect the same device directly to a suitable router LAN port using Ethernet. Verify that Ethernet is the active connection and temporarily disconnect Wi-Fi so the intended path is unambiguous.

Start with one run in each condition. If the difference is large and your data allowance permits, collect three runs per condition with a short pause between them. Alternate conditions in a second round where practical: far, near, wired, then repeat. This reduces the chance that all distant-room tests coincide with a temporary network slowdown. Record the order instead of treating measurements from different moments as simultaneous.

Do not quietly change another setting during this sequence. Changing the wireless band, stopping a VPN, and moving rooms together makes any improvement difficult to explain. If you notice a backup starting, note the time and keep that observation separate. The result may still describe real everyday performance, but it is no longer the same quiet baseline as the other runs.

6. Check that the Ethernet baseline is actually suitable

×

Ethernet is a comparison path, not an automatic guarantee of maximum performance. Inspect the cable connections and use a router LAN port rather than the internet-facing port. Microsoft recommends checking the connections and trying another available cable when troubleshooting Ethernet. Avoid bypassing the router or connecting directly to an unfamiliar modem as a casual experiment: configuration and security arrangements can differ, and doing so is unnecessary for the first comparison.

Check the negotiated link speed in your device's network information if it is available. A link reported as 100 Mbps cannot carry a 500 Mbps internet measurement through that link. A 1 Gbps link, however, does not guarantee a 1 Gbps test result. USB adapters, docking stations, intermediate switches, cabling, and router ports belong in the path diagram. Write down their stated capabilities rather than inferring them from appearance.

If the wired result is unexpectedly lower than nearby Wi-Fi, investigate the baseline before concluding that wireless is superior in every situation. Try an existing known-working cable or another suitable LAN port, making one change at a time. Do not force speed or duplex settings simply to make the displayed number larger. Use the equipment documentation or ask support if the negotiated connection does not match its expected capability.

7. Compare wireless bands and placement as separate questions

×

Microsoft's home-layout guidance describes the practical tradeoff between wireless bands: 2.4 GHz often reaches farther, while 5 GHz and 6 GHz can provide higher throughput under suitable conditions but have different coverage and compatibility limits. The highest frequency is not automatically the best choice in every room. Your device, access point, walls, and local radio environment all influence the result. Use measurements in the actual room rather than a rule that one band always wins.

After completing the initial baseline, compare bands only if your existing equipment makes that choice clear and safe. Keep the device in the same location, record which band is confirmed, and repeat the same procedure. If the router automatically chooses the band and does not expose an obvious control, do not guess. Mark the band unknown and continue with location-based evidence.

A placement experiment can be equally useful. If practical, move the router away from an enclosed cabinet or a large metal obstruction while preserving ventilation and safe cable routing. Photograph or describe the old position, change only the position, and compare again. Do not obstruct walkways with temporary cables or move equipment you are not authorized to manage. Keep the improvement only if it is repeatable and the normal activity also benefits.

8. Treat mesh nodes and repeaters as additional links

×

A strong signal beside a mesh satellite describes the nearby wireless link, not necessarily the satellite's connection back to the main router. That connecting path is commonly called backhaul. It may be wired or wireless. TP-Link's Ethernet-backhaul documentation illustrates how supported mesh units can exchange traffic over an Ethernet connection instead of relying only on wireless backhaul. Exact capabilities and supported wiring depend on the equipment; do not copy a topology from another product without checking its manual.

For diagnosis, draw a simple path in words: laptop, office satellite, main router, internet. Then compare the laptop near the main router with the laptop near the satellite. If the main-router position performs consistently better, investigate the satellite placement and backhaul before assuming your internet plan is inadequate. A repeater located where the original signal is already poor may present a visible network name without solving the upstream weakness.

If the management app shows the connected node or backhaul type, record it. A device may remain attached to a different access point after you move; allow the connection to settle and verify what you can. An Ethernet connection into a satellite is not the same as Ethernet directly into the main router when the satellite itself uses a wireless backhaul. Label the entire path honestly in the comparison.

9. A worked example: isolate a room-specific limitation

×

This is a hypothetical example for learning, not a customer case or a benchmark. One laptop, one browser, and one test method are used with other transfers paused and no VPN. At the office desk, three downloads are 22, 25, and 24 Mbps; uploads are 8, 9, and 8 Mbps; ping values are 42, 46, and 44 ms. The medians are therefore 24 Mbps download, 8 Mbps upload, and 44 ms ping.

Near the main router, downloads are 138, 145, and 141 Mbps, uploads are 18, 19, and 18 Mbps, and ping values are 24, 23, and 25 ms. The medians are 141 Mbps, 18 Mbps, and 24 ms. Direct Ethernet gives downloads of 176, 182, and 179 Mbps, uploads of 19, 20, and 19 Mbps, and ping values of 21, 22, and 21 ms. Those medians are 179 Mbps, 19 Mbps, and 21 ms.

The distant-room download median is about 17 percent of the nearby median: 24 divided by 141, multiplied by 100. This describes a large location-associated difference in this example. It is not a universal Wi-Fi efficiency threshold. Nearby wireless is much closer to the wired baseline, so a room-specific wireless investigation is more justified than immediately requesting a faster plan.

10. Use the example to choose a follow-up, not a verdict

×

In the hypothetical example, moving the laptop changed the observation substantially. Several explanations remain possible: the radio path, interference, a different access point, or an automatic band change. The next useful action is to record the connected band and node if possible, then repeat one controlled change in the office. Moving an existing access point or comparing an available band is more informative than replacing several components at once.

Suppose a permitted placement change produces office downloads of 88, 91, and 89 Mbps. The new median is 89 Mbps. Compared with the previous office median of 24 Mbps, that is an increase of 65 Mbps, or about 271 percent relative to the old value. Report both the absolute values and the percentage, because a large percentage can sound more impressive than the practical benefit warrants. This remains a hypothetical before-and-after example.

Now return to the original task. Does a normal call remain stable at that desk? If it does, save the configuration and stop changing things. If it does not, the increased throughput has not solved the whole problem. Investigate the affected application, delay variation, and time-dependent behavior. A successful diagnosis includes a useful next action and a check of the original symptom, not merely a green number.

11. Summarize repetition without hiding awkward results

×

With three comparable results, sort them and take the middle one as the median. For 92, 95, and 180 Mbps, the median is 95 Mbps, not 180. Keep the lowest and highest values too. A summary such as median 95 Mbps, range 92 to 180, three runs is more informative than reporting only the fastest attempt. The unusual result deserves a note, not automatic deletion or automatic acceptance as the connection's permanent capability.

Use a separate summary for each condition. Combining wired, nearby wireless, and distant wireless results into one average erases the distinction the experiment was designed to investigate. Similarly, do not mix morning and evening observations if time of day is the question. Include the number of runs so a single observation does not acquire the apparent authority of a longer investigation.

Three runs are a practical troubleshooting sample, not a statistically representative survey of a provider. If two conditions overlap substantially or fluctuate widely, label the conclusion uncertain and repeat later rather than choosing a winner by a tiny difference. Keep contextual notes: a network switch, an incoming call, a paused tab, or a background update can explain why one run is not comparable. There is no universal percentage gap that proves a particular fault.

12. Separate busy-hour patterns from local radio problems

×

If both nearby Wi-Fi and direct Ethernet seem slow, repeat a small comparable set at another time, preferably when the connection is otherwise quiet. Keep the same device and method. Write down household activity instead of assuming that evening slowdown must come from the provider. Local backups, shared streaming, the router, and networks beyond your home can all change at particular times.

An illustrative pattern might be a wired download median of 180 Mbps in the morning and 55 Mbps in the evening, with no deliberate changes. That pattern supports further time-based investigation, but it does not identify congestion at a specific location. Check whether the pattern repeats on another day and whether another authorized device observes something similar. These numbers are examples only, not a prediction or a claim about a particular provider.

Avoid testing continuously all evening to obtain proof. A few well-recorded observations alongside the actual symptom usually make a clearer support report than hundreds of unexplained screenshots. If the problem only affects one service while comparable measurements remain stable, record the service and time. The useful question then becomes why that particular path or application behaves differently, rather than whether the whole connection is universally slow.

13. Account for VPNs, browsers, and the device itself

×

A VPN changes the path used for some or all traffic, depending on its configuration. Record whether it is active and the selected region if known. If your policy allows it, compare the same connection with and without the VPN in a separate experiment. Do not disable an employer's required VPN or security controls. A difference identifies a path-dependent observation; it does not prove that encryption alone caused the change.

A browser test also depends on the device running it. Compare a second supported browser only after finishing the connection comparison, and label that as a browser check. Keep extensions and background workloads in mind, but do not disable antivirus or protective software as a routine shortcut. If a particular browser fails while another completes, preserve the failure message and investigate compatibility rather than treating failure as a valid zero-speed sample.

A second device can help separate device-specific behavior from a shared network problem. However, it is a follow-up comparison, not a replacement for the same-device baseline. Different wireless adapters and capabilities may explain different results. State exactly what changed. If your only available device cannot use Ethernet, say so in the final notes; a limitation disclosed clearly is better evidence than an invented wired benchmark.

14. Follow a decision path that preserves uncertainty

×

Use the following branches as a way to choose the next check. They are diagnostic prompts, not automatic fault labels. Before entering any branch, confirm that the compared runs used the intended connections and broadly similar conditions. If a measurement failed or the network changed mid-test, repeat only that condition when safe and affordable.

  • Distant Wi-Fi is repeatedly worse, while nearby Wi-Fi and Ethernet are similar: investigate the distant room, connected band, access point, and obstacles. Recheck the original activity after a single placement or connection change.
  • Nearby and distant Wi-Fi are both worse than suitable Ethernet: investigate wireless configuration, the device's wireless capability, and shared radio conditions. Compare another authorized wireless device before attributing everything to the router.
  • Ethernet is unexpectedly worse than nearby Wi-Fi: inspect its negotiated link speed, cable, port, adapter, and any intermediate equipment. A constrained wired baseline cannot validate a faster plan.
  • All three conditions are similarly slow: verify background transfers and VPN state, then compare another time or device. Shared slowness narrows the investigation but does not establish a provider fault on its own.
  • Numbers look consistent but one application still fails: document that application's symptoms and destination-specific behavior. Do not keep increasing download throughput as a substitute for investigating the actual failure.

15. Know when to stop, and what the test cannot establish

×

Speed tests use real data. On a capped or metered connection, check your allowance before repeated runs and use the smallest useful experiment. Data use varies with the measurement configuration and connection; a 30-second interface does not guarantee a fixed number of megabytes. Stop if you approach your allowance, disrupt another person's task, or already have enough evidence to select a practical next step.

This browser comparison does not identify every nearby wireless network, measure every radio channel, certify cabling, or establish contractual compliance. It cannot prove a provider is dishonest. Country-level provider averages are not a substitute for measurements at your address, and this page's worked examples are not real provider results. A local improvement also does not guarantee identical performance tomorrow or to every internet destination.

Avoid factory resets, unverified optimizer downloads, and broad configuration changes simply because a number is disappointing. A factory reset can remove information needed to reconnect. When a configuration change is justified, record the original setting and use the manufacturer's instructions so it can be reversed. Stop experimenting when the normal task works reliably, or when the remaining question requires the equipment owner or provider to inspect information you cannot access.

16. Send support a concise evidence package

×

A useful support report tells another person what happened, what was controlled, and what remains unknown. Start with the original symptom, affected location, date range, device, and connection. Include your plan's stated download and upload information exactly as described in your account documents; do not substitute an advertisement or assume that both directions have the same allowance. Explain whether the connection is fixed broadband or a mobile hotspot.

Then summarize each condition with its run count, median, and range. State the test method and recorded endpoint, VPN state, quiet or busy conditions, and Ethernet link speed if available. Attach only relevant results. Mask unnecessary public IP addresses, account identifiers, home addresses, wireless passwords, and personal details before sharing screenshots outside the provider's authenticated support channel.

An adaptable message is: I experience [symptom] at [location and times]. Using [device and method], I compared [conditions]. My wired results were [values], nearby Wi-Fi [values], and usual-room Wi-Fi [values], with [number] runs per condition. I changed [one item] and observed [outcome]. I could not verify [unknown]. Please advise which next check would distinguish [remaining possibilities]. This gives support a reproducible starting point without presenting a browser test as a legal or engineering certification.

References and further reading

×

Overview

A slow result does not tell you which part of a connection needs attention. This practical guide turns a speed test into a small, repeatable investigation: compare your usual room, a position near the router, and a direct Ethernet connection while keeping other conditions consistent. You will learn how to record evidence, interpret differences without blaming the wrong component, and choose the next useful check. All numerical case studies below are explicitly hypothetical teaching examples, not measurements collected from customers or promises about your connection.

Three connection conditions
A: distant Wi-Fi. B: Wi-Fi near the router. C: Ethernet. Keep the device and test method consistent; this illustration does not establish which condition will be faster.

ContactMethodologyWiki

Useful, at a glance

Relevant next steps to help you put this page into practice.