What is loaded latency? Idle, download and upload ping explained

What does loaded latency mean?

Latency, or ping, is the time a small piece of data takes to reach a server and come back. Loaded latency is that same round trip, measured while the connection is carrying as much traffic as it can. Netflix describes it on Fast.com as round-trip time while data-heavy applications are in use, and says the loaded and unloaded values should ideally be close.

You'll see the same idea under several names. They aren't always measured in the same way, but they answer the same question: what happens to your ping when the line is busy?

• Loaded latency, or latency under load: the plain term, used by Fast.com and most networking guides.
• Download ping and upload ping: the names in the Speedtest by Ookla apps, measured during each phase of the test.
• Working latency: the term used in networking standards work, where the idea is a network under normal use rather than an empty one.
• Responsiveness, measured in RPM: Apple's version, covered further down.
• Bufferbloat: the usual cause of a big gap. Loaded latency is the measurement, and bufferbloat is one thing it can reveal.

A classic speed test only showed idle ping, which is why a line could look perfect and still lag. Ookla added ping during the download and upload phases to its Speedtest apps in 2022, and Netflix added loaded latency to Fast.com in 2018.

What is the difference between idle and loaded latency?

Idle latency shows the best your connection can do: the distance to the test server, plus the delay in your own equipment and the route between them. Loaded latency shows what happens at home on a normal evening, when someone is streaming, a console is updating or photos are backing up to the cloud.

The two numbers point at different problems. High idle latency means the path is long or slow, so a queue isn't the cause. Low idle latency with high loaded latency means the path is fine until it gets busy. That second pattern is why a person can have a 500 Mbps plan and still lag in a match the moment a family member starts a download.

What is a good loaded latency?

Judge the rise, not the raw number. A loaded latency of 60 ms is excellent if your idle ping is 55 ms, and much less so if your idle ping is 10 ms. The Net Fix (thenetfix.com) grades the rise on the widely used Waveform bufferbloat scale, so you can compare results between the two.

There's no official standard for these bands, and sites disagree on cutoffs. They're useful because they tie a number to what you'll feel. If you only remember one figure, aim for a rise under 30 ms.

How do I read idle, download and upload ping?

Take your idle ping away from each of the other two. Then look at which direction rose, because that tells you what will trigger lag at home. These examples use made-up results to show the patterns.

Upload is often the worse of the two on US cable plans, because upload capacity is much smaller than download, so a phone backing up photos can fill it. Here's the arithmetic: 1 MB of data waiting in a buffer takes about 400 ms to drain at 20 Mbps, and every game packet behind it waits as well. The same 1 MB takes only about 8 ms at 1 Gbps. That's why the plan's headline number tells you so little about lag.

Spikes matter too. A good average with big jumps points to jitter, not a queue alone, and our jitter guide covers how to tell them apart.

Where can I test loaded latency?

Several free tools show it, each with its own labels. They measure in different ways and use different servers, so compare a tool only with itself: run the same test before and after a change.

For a reliable reading, pause other devices, use a cable if you can, and run the test twice. The steps are in how to test for bufferbloat. Our Test & Optimize runs the same check along with Wi-Fi, spikes and DNS, then lists the fixes that apply to you.

What is RPM, and how does it relate to loaded latency?

RPM stands for round-trips per minute. Apple uses it in its Network Responsiveness test, the networkQuality command on macOS, and reports it as low, medium or high. It's loaded latency turned upside down so that a higher number is better. As rough arithmetic, 60,000 divided by the delay in milliseconds gives the RPM, so 100 ms is about 600 RPM and 30 ms is about 2,000 RPM.

Apple's actual test counts several kinds of round trip, so the sum won't match its output exactly. Use the conversion to get a feel for the number, and use Apple's low, medium and high labels for the verdict. Apple says that at low responsiveness, apps such as FaceTime and games may be unreliable when other devices are busy.

Why is my loaded latency so high?

Usually because data is waiting in a long queue at the slowest point on your line, which is often your router or modem. When a download or upload fills the link, extra packets pile up in a buffer, and a game's small packets join the back of that line. It's fixable at home in most cases, so it doesn't prove your provider is at fault.

• A router with no queue control. This is the most common cause, and it's the one Smart Queue fixes.
• Wi-Fi. A weak link adds its own delay when busy, so retest on a cable before changing settings.
• A router that's too slow for the plan. Smart Queue runs on the router's processor, which can be the limit on fast lines.
• Busy hours on the provider's network. If the rise only shows up in the evening on a cable connection, see why internet slows down at night.

How do I lower loaded latency?

Turn on Smart Queue (SQM) on your router and set its limits just under your real speeds. The router then keeps the queue short and lets small packets through first. The full walkthrough is in the bufferbloat fix. In short:

1. Note your idle, download and upload ping, plus your speeds.
2. Find your router in our router guides, which give the exact setting for each brand.
3. Set limits at about 90% of your measured speeds. The Smart Queue assistant works them out.
4. Test again. If the rise is still large, lower the limits to 85% and repeat.
5. Keep the change only if the grade improved. Priority-only QoS may do little, and on some routers it can do harm, as our TP-Link guide explains.