How to Test Local Network Speed Between Two Computers

A local network speed test transfers traffic between two computers on your own network. An internet speed test transfers traffic to a remote service. Comparing them can reveal whether a problem deserves attention inside the home network, but the numbers are not interchangeable: the endpoints, software and route are different.

Use this procedure when a wired or wireless computer struggles with local transfers and you want a clearer comparison. If your immediate question is whether an Ethernet link negotiated at the expected rate, start with the slow Ethernet connection guide. A local throughput experiment answers a separate question: what can these two endpoints transfer under these conditions?

Choose the endpoints before the tool

Write down the two computers, their connection types and the path between them. “Laptop on office Wi-Fi to desktop on router Ethernet” is a useful description. “My LAN” is too vague if the path includes mesh backhaul, a switch, a dock or a second router. Include those components without assuming which one limits the transfer.

Use equipment and a network you control. Do not expose a test server through port forwarding or use someone else’s public endpoint for this local experiment. Arrange a quiet period because the test itself transfers traffic and can compete with normal use. Stop if it disrupts an important call or other household activity.

A bounded iperf example

The ESnet iperf3 manual describes a client/server throughput test. On a supported installation, one endpoint listens and the other initiates the test. Check the manual installed with your version before using options. The examples below illustrate a short TCP test; they are not commands for configuring your router.

iperf3 -s -1

Run that on the receiving computer. On the other computer, replace LAN_ADDRESS with the receiving computer’s current local address:

iperf3 -c LAN_ADDRESS -t 10

The example requests ten seconds of traffic; the one-off server option limits the listening session to one client. Confirm the connection is local and permitted by your normal firewall policy. If it cannot connect, investigate address, reachability and the specific allowed application rather than disabling the firewall.

There is an important platform boundary: ESnet’s FAQ says iperf3 is not officially supported on Windows and recommends iperf2 there. Do not assume a random Windows download is an ESnet-supported build. Windows readers should use a supported tool/distribution appropriate to their environment and its own documentation; iperf2 and iperf3 instructions are not interchangeable.

Keep a small comparison record

1. Record the tool and version on both endpoints, addresses, connection types and chosen duration.
2. Observe one short run with other transfers paused. Save the actual summary and any error message; do not replace an unsuccessful result with zero.
3. Repeat under the same conditions. A second result is more useful than a single unusually high number.
4. Change one path component you can safely control, such as connecting the same laptop by Ethernet instead of Wi-Fi, and repeat the same procedure.
5. Stop the listener when finished. Keep the result labelled as a local test between the named machines.

For a room comparison, keep the receiving endpoint fixed while moving the wireless computer. Otherwise you change both ends and cannot tell which change accompanied the result. For Wi-Fi versus Ethernet, keep the computer and activity constant and document the interface actually used.

Interpret a difference without overdiagnosing it

If the local test improves after one controlled change, you have a lead about that setup. You have not proved that every other component is healthy. Both endpoint machines still participate in the measurement, and a software, device or path limitation can influence what you observe.

If local traffic looks satisfactory but internet traffic remains slow, compare a separate internet speed test under similar conditions. That shifts attention toward the additional remote path or service, but does not automatically identify the ISP as responsible. The browser and iperf experiments use different methods, so compare patterns rather than demanding identical numbers.

If both tests are slow, start with what they share: the test computer, interface and nearby network path. If a local file copy is slow while the network-only test is better, keep storage and file-serving behaviour in the investigation. Our NAS home-server speed guide addresses that different task.

Know what would make the experiment inconclusive

Changing computers, Wi-Fi locations, test directions and background activity together produces a result with too many explanations. A failed connection also needs its own troubleshooting; it is not a throughput measurement. Keep those cases out of an average.

Data volume matters even on a local network. As arithmetic only, a sustained 100 Mbps for ten seconds represents about 125 MB before overhead. Faster rates move more data in the same time. This is an estimate, not a recorded result from your equipment. When seeking help, provide the topology, versions, bounded commands and summaries so someone can reproduce the question you actually tested.