How to Measure Latency: Tools, Commands and What Results Mean

To measure latency, send small packets to a server and time how long each takes to come back. That is round-trip time, shown in milliseconds. The quickest ways are a browser speed test that reports latency, or the ping command on your computer. For a fuller picture, add traceroute to see where delay builds up, and run the test more than once so you catch the swings as well as the average.

This guide covers each method, what each one really measures, which numbers are good, and what to do when they are bad.

TL;DR

• Latency is the time a packet needs to travel to a server and back, measured in milliseconds. It is separate from download and upload speed.
• ping gives you round-trip time and packet loss. traceroute shows delay at each hop on the path.
• ICMP ping is simple but can mislead. TCP-based tests are closer to how real apps behave.
• One number is not enough. Test several times, at different hours, and look at jitter and latency under load too.
• You can run a browser test on any phone or laptop with our speed test, which reports latency, jitter, packet loss, bufferbloat and DNS time together.

Table of Contents

• What latency actually measures
• Method 1: A browser test (fastest, no install)
• Method 2: Ping from the command line
• Method 3: Traceroute to find where delay builds
• Method 4: TCP and browser-level measurements
• Comparing latency test tools
• Why latency under load matters
• What counts as good latency?
• What causes high or unstable latency
• A simple troubleshooting order
• Our perspective
• Check your connection with GrandSpeed
• FAQ
• Sources

What latency actually measures

MDN describes network latency as the time it takes for a request to travel from your computer to the responding computer and for the reply to come back, and it is generally measured as round-trip delay. That is the number your ping tool prints.

The ITU defines delay a little more formally: the interval from the first bit entering the network until it arrives at the user. It includes transmission, propagation and node-processing time. Distance sets a floor you cannot beat. Everything above that floor comes from the equipment and queues in between.

Latency versus speed, jitter and packet loss

These four numbers get mixed up constantly, so it helps to pin them down:

AWS draws the line between capacity and delivery this way: bandwidth is the theoretical maximum, while throughput is what actually passes over a given period. Our short piece on bandwidth versus throughput goes deeper. A 1 Gbps plan can still feel sluggish if each packet waits 150 ms.

Jitter, according to the ITU, is the variation between maximum and minimum delay in a time window, driven mostly by queuing. One caution from BITAG: there are several different industry definitions of jitter, so check how a tool calculates it before comparing two tools.

Network latency is also not the same as the lag you feel inside a game or an app. On-screen delay can come from rendering, capture and overlays on your own machine. If your ping is low but input still feels late, read about fixing overlay input lag on PC before blaming your provider.

Method 1: A browser test (fastest, no install)

A web-based test works on a phone, tablet or laptop, over Wi-Fi or cellular. That makes it the right first step for most people.

1. Close streaming, downloads and cloud backups. Anything using the connection will inflate the result.
2. Connect the way you normally do. If you play on Ethernet, test on Ethernet.
3. Open GrandSpeed and run the test.
4. Read latency and jitter first, then packet loss and bufferbloat, then speed.
5. Repeat two or three times, then again at a busy hour such as the evening.

A browser test measures the path from your device to a test server, so it reflects your home network plus your provider. It does not tell you the delay to a specific game server or website. For that, use the command line.

Method 2: Ping from the command line

Cisco explains that ping sends ICMP Echo messages to an address and reports reachability, round-trip delay and packet loss. It is built into Windows, macOS and Linux.

Windows

Open Command Prompt and type ping google.com. Windows sends four packets by default. To keep it running and watch the trend, use ping -t google.com. On macOS and Linux, ping runs until you stop it with Ctrl+C.

Reading the output

You will see a line per packet with a time= value in milliseconds, then a summary with minimum, average and maximum, plus the percentage lost. Look at the spread, not only the average. A result of 18 ms minimum and 240 ms maximum is a problem even if the average looks fine.

What to ping

• Your router (often 192.168.0.1 or 192.168.1.1): this isolates your home network. Anything above a few milliseconds on Ethernet, or erratic on Wi-Fi, points to your local setup.
• A public DNS server such as 8.8.8.8: this tests the path through your provider to the wider internet.
• The service you care about: a game server address, a video platform, or your office VPN.

Pinging the router first, then outward, is the cheapest way to split "my Wi-Fi" from "my ISP".

The limits of ICMP ping

Ping is simple, but routers and firewalls may treat ICMP differently from regular traffic, and some hosts deprioritize or block it. Google Cloud notes that TCP-based tools such as netperf represent real-world applications more closely than ICMP ping. A ping result is a useful clue, not the final word.

Method 3: Traceroute to find where delay builds

Traceroute, per the same Cisco guide, discovers the route packets take and the round-trip time to each hop. Use tracert example.com on Windows and traceroute example.com on macOS or Linux.

How to read it:

• Each row is a hop, usually a router. Three times are shown per hop.
• A big jump that persists in every later hop marks where the delay begins. Often it is the step where traffic leaves your provider's network, or crosses an ocean.
• A single slow hop followed by normal hops is usually a router that is slow to answer probes, not a real bottleneck.
• Stars (* * *) mean no reply to that probe. A few are normal. Stars through to the end may mean a firewall, or a genuine break.

Save the output when you contact support. It gives a provider something concrete to check.

Method 4: TCP and browser-level measurements

Web apps do not run on ICMP. They open TCP connections, so measuring the handshake gets closer to what your browser or game client experiences.

• TCP ping. APNIC describes using a TCP traceroute to a web server's address on port 80 to mimic an HTTP connection's handshake, and found that the results are similar to ICMP round-trip times in many cases.
• Browser timing. Developers can read connectStart and connectEnd from the Resource Timing API to see how long the TCP handshake takes for any request a page makes.
• netperf. A command-line tool that runs latency and throughput tests over TCP, useful for servers and lab work rather than home troubleshooting.
• Passive measurement on live traffic. Network engineers use methods such as the Alternate-Marking Method in RFC 9341 to measure packet loss, delay and jitter on real traffic. It is for carriers and large networks, not for home users.

Comparing latency test tools

There is no single best latency test. The right one depends on the question.

Ookla's Speedtest is the best-known consumer option. As of August 2026, one review lists a free tier, a $0.99 one-time fee to remove ads, and a $4.99 monthly subscription for its VPN product. A basic speed test is fine for a first look. Where we differ is scope: we add bufferbloat, DNS resolution time and streaming-specific checks, which a speed-only test skips. Our write-up of speed test alternatives that catch jitter and packet loss explains why that matters.

Why latency under load matters

An idle ping can look great while your connection falls apart the moment someone uploads a video. That is bufferbloat: queues in the router or modem fill up and every packet waits behind them. The ITU notes that processing and queuing delay vary with traffic, packet size and scheduling, which is exactly why loaded latency differs from idle latency.

To check it yourself, ping a stable host with ping -t while you start a large upload or a 4K stream on another device. If the times jump from 20 ms to several hundred, you have a queuing problem, and the fix is usually in the router (smart queue management, or a better router), not a faster plan. Our guide to the best ping test and RPM under load shows how to measure it more rigorously.

What counts as good latency?

Thresholds are rules of thumb, not standards. One US provider's guide treats under 20 ms as excellent, 20 to 40 ms as good, and anything over 100 ms as noticeably laggy for gaming, calls and streaming. Use these as a practical guide:

The comfortable and warning ranges in the second table are our working guidance, drawn from how these applications behave, rather than a published standard. Consistency often matters more than the headline figure. A steady 60 ms feels better in a game than 20 ms that spikes to 200 ms.

On the question people often ask, whether 177 ms is good: for most interactive uses, no. It is workable for browsing and video-on-demand, but gaming and calls will feel delayed. If you see it to a nearby server, something is wrong. If you see it to a server on another continent, it may simply be distance.

What causes high or unstable latency

The ITU lists the contributors: distance-related propagation delay, transmission delay, processing delay, queuing, congestion-related drops, signal degradation and hardware faults. Translated to a home setup:

• Weak Wi-Fi. Distance, walls and crowded channels add delay and jitter. Test on Ethernet to confirm.
• Congestion. Evening slowdowns on a shared cable or cellular network.
• Bufferbloat. Latency that climbs only when the line is busy.
• Distance to the server. Physical limits. Choose a closer game region.
• Faulty or old hardware. Failing cables, aging routers, noisy lines.
• Packet loss. Retransmissions add delay. See why packet loss happens and how to prove it to your ISP.

A simple troubleshooting order

1. Test on Ethernet next to the router. If it is good, the problem is Wi-Fi.
2. Ping your router, then 8.8.8.8, then your target. The first big jump shows where the delay enters.
3. Run the test while the house is idle, then while it is busy, to expose bufferbloat.
4. Restart the modem and router, and update the router firmware.
5. Run tests at three times of day and keep the results. Our test history guide explains how to track them.
6. If the problem persists on a wired connection, contact your provider with your ping, traceroute and packet loss figures. For calls, the steps in fixing VoIP jitter help you gather them.

Our perspective

Most people who come to us have the same story: they pay for a fast plan, and calls still stutter or streams still drop. Speed rarely explains it. In our experience the cause is usually delay that varies, queues that fill under load, or a little packet loss, none of which a download number shows.

So we do not treat latency as a single figure. We want it next to jitter, loss, bufferbloat and DNS time, tested on the device and network you actually use. For streamers, we go one step further: a latency reading means little until you know whether your upload can hold a given resolution steadily. That is why our tools translate results into what your connection can support for streaming and broadcasting.

Our advice: run tests repeatedly, at different times, and trust the pattern over any single reading.

— Peter D

Check your connection with GrandSpeed

If you want latency, jitter, packet loss, bufferbloat and DNS in one run, start with the GrandSpeed speed test. Streaming or broadcasting? The Livestream Upload Speed Monitor checks that your upload stays steady, and the Video Streaming Quality Test shows what resolution your connection supports. The tools are free and run in any modern browser.

FAQ

What is the best way to measure latency at home?

Start with a browser speed test that reports latency and jitter, then confirm with ping to your router and to a public server. Run both on Ethernet and on Wi-Fi. Comparing the results tells you whether the delay comes from your home network or your provider.

Is ping the same as latency?

In everyday use, yes: ping is the round-trip time most tools report. Strictly, ping is the tool, and latency is the delay it measures. Because ping uses ICMP, TCP-based tests can reflect real application behavior more accurately.

How do I measure latency to a specific website or game server?

Ping its hostname or IP address, and run traceroute to see each hop. If the host blocks ICMP, a TCP-based check on the port it uses will often work. Some game clients also show their own ping, which is the most relevant number for that game.

Why does my latency change from test to test?

Queues and processing delays vary with traffic, so some variation is normal. Large swings point to Wi-Fi interference, congestion, or bufferbloat. Test several times and at different hours, and judge the range, not one result.

Sources

• Understanding latency, MDN
• ITU-T Recommendation Y.2617
• Throughput vs latency, AWS
• Latency explained, BITAG
• Understand Ping and Traceroute Commands, Cisco
• How to see your network latency using a ping test, RoseHosting
• Using netperf and ping to measure network latency, Google Cloud
• Measuring your web server reachability with TCP Ping, APNIC
• PerformanceResourceTiming: connectStart, MDN
• RFC 9341, Alternate-Marking Method
• Ookla review (2026), Market Intelligence Tools
• What is internet latency, Optimum
• Ultralow latency overlay: fix overlay input lag on PC

Recommended

• Best Ping Test: RPM Reveals Latency Under Load
• 6 Steps to Troubleshooting Latency With Per Hop Isolation and GrandSpeed
• Ping vs Latency: Why One Ping Misses Jitter in Games and Calls
• Measure DNS Latency From Your Network (Client→Resolver→Authoritative)