What Are SQM and QoS?

Quality of Service: choosing who goes first

Quality of Service, usually shortened to QoS, is a router feature that lets you set priorities among the traffic passing through it. In its simplest form, you might tell the router that a particular device, or a particular kind of traffic such as video calls, should be sent ahead of everything else whenever there is competition for the link.

Quality of Service (QoS)

A set of rules a router applies to decide which packets are sent first when more than one type of traffic is competing for the same limited link.

This can genuinely help in specific situations — for example, making sure a video call's traffic is not stuck behind someone else's large download. But basic QoS has a limitation: it decides the order things leave the queue, without necessarily doing anything about how long or deep that queue is allowed to become in the first place. If the queue itself is oversized, even prioritised traffic can still wait behind a large amount of lower-priority data that arrived just before it.

Smart Queue Management: keeping the queue itself short

Smart Queue Management, often shortened to SQM, takes a different approach. Rather than only choosing an order, it actively manages the size of the queue itself, typically by capping the router's outgoing rate slightly below the connection's true maximum and using queuing techniques that keep individual flows from monopolising the link.

Smart Queue Management (SQM)

A router feature that actively limits queue depth and shares link capacity fairly between different flows, aiming to keep delay low even while the connection is fully loaded.

The effect is that when a large download or upload is running, small and time-sensitive packets — game inputs, voice frames, page requests — do not end up stuck behind a long backlog, because the backlog itself is kept short by design. This is why SQM is closely associated with reducing the kind of delay described elsewhere as loaded latency or bufferbloat: it addresses the queue directly, rather than only reordering what is already in it.

How the two approaches differ in practice

Both features can appear under similar-sounding menu names, which causes a lot of confusion. The practical difference is what each one actually changes.

Some routers offer only one of the two, some offer both under a combined menu, and some use the terms loosely enough that the actual behaviour has to be checked by testing rather than assumed from the label.

The throughput trade-off, and why it exists

SQM generally works by capping the rate the router will send at to slightly below the link's actual maximum, often somewhere in the region of 85 to 95 percent of the measured capacity. That margin exists on purpose: it gives the router's own queue management a chance to act before the link itself becomes the bottleneck, since a queue that forms inside equipment you control is far easier to keep short than one that forms somewhere else on the path, such as inside provider equipment.

The result is that a speed test that measures top speed may show a slightly lower number with SQM enabled than without it. This is an expected and deliberate trade, not a malfunction. For most everyday use — calls, games, browsing while something else transfers — a small reduction in top speed in exchange for a much smaller amount of delay under load is a reasonable trade, though it is a genuine trade rather than a pure improvement.

Why settings differ so much between routers

Router manufacturers implement these features differently, using different underlying algorithms, different default settings, and different amounts of manual configuration. Some routers automatically detect the connection's speed and set a sensible cap; others require you to enter your download and upload speeds manually, and get poor results if those numbers are wrong or become outdated after a plan change.

This variation is exactly why there is no single set of instructions that applies to every router, and why this article does not attempt to give model-specific steps. What holds true across routers is the underlying idea — actively managing queue depth reduces delay under load — and the only reliable way to know whether a given router's implementation of that idea is working for you is to measure before and after.

When it helps, and when it does not

SQM tends to help most on connections that already show a large gap between idle latency and loaded latency, particularly when that gap appears mainly during upload. It helps least — or not at all — when the bottleneck causing delay sits somewhere the router cannot influence, such as further along a provider's network, or when the connection was already well-behaved under load to begin with.

QoS-only prioritisation tends to help in more specific situations, such as a household where one predictable type of traffic — a work video call, for example — regularly competes with another predictable type, such as a household member's gaming or streaming. It is less useful as a general-purpose fix for delay that appears across many kinds of activity at once.

Establishing a baseline and testing one change at a time

Before changing any setting, record a baseline: run a loaded-latency test with nothing else changed on the network, and note the idle figure and both loaded figures. This baseline is what any later result gets compared against, and without it, an improvement or a lack of one is just an impression rather than a measurement.

After enabling SQM or configuring QoS rules, repeat the test under similar conditions — similar time of day, similar background activity — and compare the new result with the baseline. Norynt's Verification Wizard is built for exactly this kind of before-and-after comparison, walking through the steps of running a baseline test, changing one condition, and running the comparison test.

The reason to change only one condition at a time is straightforward: if you enable SQM and also pause a background download in the same session, an improvement could be due to either change, and you will not know which setting to keep. Isolating one variable at a time is what turns a hopeful guess into a useful finding.