The speed field in a slicer is a request, not a promise. Four separate systems get a veto over it — the hotend's melt rate, the machine's acceleration, the part's geometry, and the material — and on most prints at least one of them is already saying no before you touch the number.
Understanding which one is talking is the difference between a useful change and an evening of disappointment.
Where the ceiling appears
The same 80 mm bracket used across this site: 95 cm³, three walls, 20% infill, 0.2 mm layers, a 0.4 mm nozzle. Only the requested speed changes.
| Requested speed | Flow asked for | Time | What happened |
|---|---|---|---|
| 80 mm/s | 7.2 mm³/s | 5.827 h | comfortably inside the melt rate |
| 150 mm/s | 13.5 mm³/s | 3.16 h | still inside it, just |
| 300 mm/s | 27 mm³/s | 2.605 h | flow limited — the head averaged 183.3 mm/s |
Between the second row and the third the requested speed doubles and the time falls by about a sixth, because the hotend cannot melt fast enough to use the rest. Ask the print time estimator for 300 mm/s and it answers with 183.3 — the speed you would actually get is the number worth putting in a profile, not the one you typed.
Doubling from 80 to 150 nearly halved the print. Doubling again from 150 to 300 removed a little over half an hour, because the hotend could not melt plastic fast enough to feed it. Everything above the ceiling is a number in a profile that the machine politely ignores.
Volumetric flow is the real unit
A nozzle lays down a cross-section — line width multiplied by layer height — and moves along at some speed. Multiply the two and you get a volume per second, and that is what the hotend has to melt.
This is why the same speed setting means completely different things in different profiles. At 0.2 mm layers and a 0.45 mm line width, 150 mm/s asks for 13.5 mm³/s. At 0.3 mm layers through a 0.6 mm nozzle, the same 150 mm/s asks for well over double that, and a hotend that coped with the first profile will not cope with the second.
If you take one habit from this page, make it this: when you change layer height, nozzle or speed, ask what flow you are requesting, not what speed. Choosing a nozzle size works the same relationship from the other end.
The four limits, in the order they usually bite
1 — Acceleration and geometry. On any part with detail, the head spends most of its time speeding up and slowing down and never reaches the commanded speed at all. A part made of short segments prints at the same wall-clock time whether the profile says 100 or 250 mm/s. This is the limit that catches people first, and it is why a slicer's estimate for a complex part barely moves when you raise the speed.
2 — Melt rate. The case above. It bites on large simple parts, big nozzles and thick layers — exactly the jobs where you were hoping for a win.
3 — Ringing and surface quality. Raise speed and the accelerations that ring the frame get larger. Ringing and ghosting is the symptom, and input shaping mitigates it rather than removing it.
4 — The material. TPU cannot be pushed fast through an extruder whatever the hotend can melt. Filled materials abrade faster at speed. PETG's surface degrades at speeds PLA tolerates.
Which knob is worth turning
If a print is too slow, these are the changes ranked by effect per unit of effort:
- Layer height. The most effective single change, and it moves time nearly inversely — see layer height and strength for what it does and does not cost you.
- Nozzle diameter. Big win on functional parts, and it raises the flow ceiling as well as the volume laid per pass.
- Wall and infill settings. Removing material you did not need is faster than laying it quickly.
- Speed itself, fourth, and only after checking whether you are already flow limited.
- A high-flow hotend, if you have done the first four and are still capped.
Notice that three of the five reduce how much work there is, and only two make the work happen faster. That ordering is the practical content of this page.
Why the machine takes longer than the slicer said
Every slicer estimate is optimistic by some margin, and the reasons are systematic rather than random:
- Acceleration and jerk limits are modelled approximately, and the shorter the moves the worse the approximation gets.
- Heat-up, bed levelling, purge lines and tool changes are often excluded or under-counted.
- Retractions, Z hops and travel moves accumulate; on a plate of many small parts they can dominate.
- Firmware may cap speeds below what the profile requested, silently.
The estimator on this site states an error band of plus or minus 30% for exactly this reason. Treat any single-figure prediction as a bracket, and if you need a real number for quoting, time one print and calibrate your expectations against it — which is the same discipline pricing printed parts depends on.
The speeds inside the profile that are not the speed
One number gets adjusted and half a dozen others quietly do not. Most slicers hold separate speeds for outer walls, inner walls, infill, top and bottom surfaces, bridges, supports and travel, and several of them are expressed as a percentage of the main figure rather than as absolute values.
The consequences are worth knowing. Outer wall speed is the one that decides surface quality, and it is usually set well below the headline figure — so raising the headline figure changes the infill and leaves the visible surface exactly as it was. Bridge and overhang speeds are deliberately slow and should stay that way. Travel speed costs nothing in quality and is frequently left far below what the machine can do, which on a plate of many small parts is a real and free saving.
If you are going to tune speeds at all, tune infill and travel upward first. They are the two that carry the least risk and, on most geometry, the most time.
A speed policy worth keeping
Set a conservative baseline profile you trust and leave it alone. Make a second profile for functional parts with a bigger nozzle and thicker layers, which is where the real time lives. Reserve high speeds for simple, chunky geometry where the machine can actually reach them, and accept that a detailed model prints at the speed its corners allow no matter what you type.
The fastest print is not the one with the biggest number in the speed box. It is the one that asked for less work.