Feedbench
Speeds and feeds for CNC routers

Every other calculator models the cutter. Feedbench models your machine.

Tell it what you actually own. It works out which single limit is binding your cut — power, belt stretch, drive force, tool deflection, the feed ceiling — and gives you advice that fits how your machine fails, not generic advice that makes it worse.

Open the calculator How it works

Free forever. No account, no email, nothing you enter leaves your browser.

Worked example
Shapeoko 3 · Makita RT0701C · hardwood · 12.7 mm 2-flute upcut · full-width slot
Machine feed rate 104%At the limit
Wants 2,600 mm/min; the Shapeoko 3 tops out at 2,500 — our estimate, because Carbide 3D publish no feed rate.
Spindle power 100%At the limit
288 W needed, 288 W available at 10,000 RPM.
Belt stretch 61%Headroom
The 9 mm belt is the softest part of this machine, and here it is not what stops you.
What set your depth
Spindle power
Depth was shed to 2.49 mm per pass before every depth-sensitive limit was satisfied. More frame will not help — you are out of watts, not out of stiffness.
Real output from the current model. Figures are estimates, not measurements.
34
machines
32
spindles
21
materials
16
machine architectures
47
named cutters, 7 suppliers
Why it is different

A number without a reason is just a guess you can’t check

It names what is binding

Nine constraints are scored against your setup at once. Feedbench shows you which one actually ran out first, what it ran out of, and by how much. Two machines with the same feed rate can be limited by completely different things, and that difference is the whole point.

The advice fits the architecture

A belt machine and a ballscrew machine do not fail the same way. Told you are belt-limited, dropping the feed while keeping full-width engagement makes it worse: you keep the force and lose the chip. Feedbench knows which of the sixteen architectures you are on and says so.

It knows what it doesn’t know

Where two suppliers publish figures that contradict each other, both are kept and labelled rather than averaged into a number nobody published. Every derived value is marked as an estimate in the interface.

The test grid

Stop calculating your limit. Cut it.

Feedbench writes a G-code program that cuts a matrix of feed rates against depths of cut in a piece of scrap. You run it, you look at it, you listen to it. Your machine reports its own real limit instead of having one calculated for it — wear, tram, spindle age and all.

This is the part no chart can do for you. A rigidity class is an engineering judgement about a model of your machine. A test grid is your machine, this week, with the bit in the collet right now.

The app sizes the cells so they stay cuttable, keeps the whole program inside your work area, and refuses to emit anything it cannot post correctly.

( FEEDBENCH TEST GRID )
( FEED 1200-3000 mm/min  DOC 1.0-4.0 mm )
G21 G90 G17
M3 S18000
G4 P3.0            ( spin-up dwell )
G0 X0.000 Y0.000 Z5.000
( CELL 1  F1200  DOC 1.0 )
G1 Z-1.000 F300.0
G1 X60.000 F1200.0
G0 Z5.000
( CELL 2  F1800  DOC 1.0 )
G1 Z-1.000 F300.0
G1 X60.000 F1800.0
...
Free forever, and here is why that is credible

Feedbench’s most useful content documents where tooling vendors’ own published figures contradict each other. That editorial line would not survive selling those vendors banner space, so it never will. The only adverts it carries are for CADCAMCUT and PanelForge, which are ours.