Why Is My Internet Slow When Someone Is Downloading?

Here is a scene from my house. My son starts a 60 GB game download on the console, and five minutes later my video call freezes, my voice turns robotic and someone shouts "the Wi-Fi is broken" from the living room. The speed test still shows a healthy number. So what is going on?

In my performance testing days, this was the classic pattern we looked for in load tests: throughput holds steady while response times shoot up. The server is not slower, it is just busy, and every new request waits in a queue. Your home internet behaves exactly the same way. In this article, let us look at why it happens, how to measure it with free tools, and what actually fixes it, using only figures the vendors themselves publish.

Why does my internet slow down when someone downloads?

Think of your connection as a single checkout lane at a supermarket. Mbps is how fast the cashier scans items. Latency is how long you wait from joining the lane to walking out. If one shopper arrives with three overflowing trolleys, the cashier keeps scanning at the same speed, but you, holding a single bottle of milk, now wait behind all of it.

That is a download hogging the line. Files, game updates, cloud photo backups and big email attachments send huge amounts of data as fast as the connection allows. Routers and modems hold the excess in a buffer, a waiting line for packets. When that buffer is large and full, the tiny, time-sensitive packets from a video call, a game or a web page sit in the same queue. People often call this bufferbloat, and the measurement that reveals it is called loaded latency.

Uploads are often worse than downloads, because many home plans have much less upload capacity, so the queue fills sooner. If you are curious why the two numbers differ, I covered that in upload vs download speed.

What is loaded latency, and how is it different from ping?

Ping, or latency, is the round trip time for a small piece of data to reach a server and come back, measured in milliseconds. The FCC'sbroadband label glossary describes typical latency as the time it takes data to travel from one point to another, notes that it is often called lag, and says lower latency means a higher-quality connection.

The catch is when you measure it. According toCloudflare's explanation of its speed test, idle latency is measured when no data is transferring, and loaded latency is measured by sending small empty requests at regular intervals while the download and upload tests are running, so it shows the round trip time while your connection is busy. Cloudflare also reports jitter, the average variation between consecutive round trips, and packet loss, the percentage of packets that fail to reach their destination.

The difference between the two latency numbers is the useful part. Cloudflare says that if the gap approaches 100 milliseconds or more over a wired connection, your internet provider is likely at fault. On Wi-Fi, the gap can also come from your own router and the radio link, which is why testing with a cable first matters.

Why does latency hurt more than slow Mbps?

Most things you do online are conversations, not one long transfer. A web page needs many small round trips before it appears. A video call sends a steady stream of tiny audio and video packets that must arrive on time. A game sends small position updates many times a second. Cloudflare notes that video calls, streaming and gaming can be more sensitive to latency, jitter and packet loss than to bandwidth.

Here is a simple way to see it. If every request has to wait for the previous answer, the round trip time sets how many conversations you can finish in a minute. Apple uses a similar idea, Round-trips Per Minute, in its responsiveness test (more on that below). The table below is just arithmetic, 60,000 milliseconds divided by the round trip time, worked out when this page is built. It is not Apple's exact method, but it shows the trend.

Same plan, same Mbps, yet going from a quiet line to a full queue cuts the number of round trips you can finish by more than thirty times. That is why the web feels sluggish and calls stutter while the speed test number looks perfectly fine.

What do calls and games actually need?

Here are the published figures I could find for latency, jitter and packet loss, each from the vendor's own page. Notice that each one is measured to that vendor's service, not to the nearest speed test server.

Microsoft's Teams network guide also describes the symptoms you hear when this goes wrong: calls with static that cut out, or voices that sound like robots, which it says could be jitter or packet loss. If that is what your family hears every time a backup kicks in, you are looking at a queue problem, not a speed problem.

How can I test for bufferbloat at home?

I will be honest about our own tool first. The Guess the Speed test measures ping and jitter to a nearby Cloudflare server while the line is quiet, then measures download and upload. It does not measure loaded latency or packet loss, and you can read exactly what it does onhow it works. It tells you how healthy your home line and Wi-Fi are, not your latency to a game server or a meeting service. For bufferbloat, use one of these:

1. Cloudflare speed test. Cloudflare's test reports idle latency, loaded latency during download and upload, jitter and packet loss. Its Aggregated Internet Measurement page explains how those metrics are turned into quality scores for streaming, gaming and real-time communication. Compare idle latency with loaded latency.
2. Apple Network Responsiveness test. Apple's support page says it measures how well a network multitasks when several devices and apps share it, and reports Round-trips Per Minute, where higher is better. On a Mac with macOS Monterey or later, type networkQuality in Terminal. On iPhone, it lives under Developer settings. Apple groups results as Low, Medium or High, and says that at Low, someone downloading a movie or backing up photos can make FaceTime calls or gaming unreliable.
3. A do-it-yourself ping test. On Windows, theping command sends echo requests and shows the round trip time for each reply.ping /t keeps going until you stop it with Ctrl+C. Start it against your router or a well-known server, note the usual milliseconds, then start a big download on another device and watch what happens. A jump from tens to hundreds of milliseconds is bufferbloat you can see with your own eyes.

Whichever test you use, run it twice: once on Wi-Fi and once with a cable plugged into the router. If loaded latency is fine on the cable and bad on Wi-Fi, the radio link or router placement is the bottleneck. If it is bad on both, the queue is at the router or the line itself.

How do I fix lag when someone is downloading?

Start with the cheapest fix and work up. In my house, the first three solved most of it.

1. Turn on Smart Queue Management if your router has it. Apple suggests this directly: some routers offer Smart Queue Management (SQM) to keep responsiveness consistently high, though Apple notes these higher-end home gateways generally need some expert configuration. Check your router's own manual for whether it has this setting and what it is called.
2. Schedule the heavy stuff. Game updates, cloud photo backups and big uploads do not need to run at 7 pm. Many consoles and backup apps have a setting to pause, schedule or limit them, so check before blaming the plan.
3. Use a cable for the device that matters. NVIDIA recommends wired Ethernet or 5 GHz Wi-Fi for GeForce NOW, and a cable removes the Wi-Fi part of the delay entirely.
4. Turn off the VPN for calls. Microsoft's Teams guidance warns that routing traffic through a VPN can add latency and jitter.
5. Upgrade the plan only if the line is genuinely full. If several people really are streaming, calling and gaming at the same time, you may simply need more capacity. Add everyone up with the household speed calculator before you spend money. A faster plan does not shrink a badly managed queue, it just fills a bigger one.
6. Call your provider with evidence. If loaded latency stays high on a wired connection with SQM enabled, and the gap is near the 100 ms mark Cloudflare mentions, take screenshots of the idle and loaded results. The plan speed check walks through what else to collect.

Is my speed test lying to me?

No, it is answering a different question. A speed test asks "how fast can this pipe move a big file?" Lag during downloads asks "how long does a small packet wait when the pipe is full?" Both matter. A connection with plenty of Mbps and terrible loaded latency feels slower than a modest plan with a well-managed queue. When people ask me what a good internet speed is, I now always add: and check what your ping does when someone else is downloading.

Want to make it a family experiment? Play a round of Guess the Speed on a quiet line, then play again while the console downloads an update. Everyone guesses first, so you will find out who in the house really knows your Wi-Fi.

Frequently asked questions

For gaming-specific speeds and ping, read how much internet speed you need for gaming.