How to Read a CNC Machine Spec Sheet: Every Number Explained

A CNC machine spec sheet is a foreign language that happens to use numbers. Every row on it measures something specific, under specific conditions, with specific measuring equipment — and none of those conditions is “cutting your part.” Two machines with nearly identical spec sheets can behave completely differently on the floor, and the same number quoted by two builders can mean different things, because the sheet is a vocabulary you have to learn before the comparison is meaningful.

This guide is the reference for reading that vocabulary. It walks down a spec sheet field by field — what each number means, how it is measured, and where the marketing hides — and it ends with a method for comparing two sheets honestly. It is the field-level companion to our guide to choosing a machine, which gives the framework (which four specifications decide a purchase); this page explains every row in the datasheet behind them. For what the numbers cost and what is really inside the machine, see the cost guide. Definitions of the machine terms below are in the CNC glossary.

The ground rules before any number

Three facts make every row easier to read honestly:

1. A spec is a claim measured in a defined way, not a fact about your job. Most dimensional and accuracy figures come from tests on a new, empty, idle machine in controlled conditions — no cutting forces, no worn ways, no hot spindle. Treat the sheet as the machine’s best behaviour, and plan margin on top of it.
2. Advertised and usable are different numbers. Maxima on the sheet (max swing, max RPM, max rapid) are frequently not the number that describes real capability. The usable number — what you can actually cut — is usually smaller, and the sheet often hides it in a footnote or a second row.
3. Same label, different measurement. Two builders can both write “±0.005 mm accuracy” and mean different things, depending on the standard used and the test procedure. Before comparing two figures, compare the method behind them.

Read the machine type and the condition first

Everything else assumes you know what kind of machine you are reading. A machining centre and a lathe describe their envelopes, spindles and workholding in different languages, and mixing them up is the first way a sheet misleads. Note the machine family, the axis arrangement, and — often more consequential than any travel — the control and the company behind it: control brand and vintage decide software compatibility, parts availability and how easily you can run it, and the support network decides whether the machine stays running. On a used machine, the most important “specification” is not on the sheet at all: the calibration certificate, spindle-hours record and service history. Read those before the travels.

Envelope and table: where the marketing hides twice

The travels (X/Y/Z) are the numbers most buyers compare first, and they are the range of motion of the axes — not the size of part you can machine. The part must fit within the travel together with the tool, the holder and the fixture around it, so the usable envelope is smaller than the travel. Two further rows define usability and are easy to miss: on a mill, the spindle-nose-to-table distance (how tall a part and how long a tool can reach) and the spindle-centre-to-column clearance (how far from the column face you can actually work); and the table load rating, which must exceed your heaviest part plus fixture and pallet, with margin. A machine whose travels swallow your part but whose table sags under your fixture is not big enough.

On a lathe the trap is built into the language. Swing over bed is what the machine can physically clear — the marketing number. Swing over the carriage (cross slide) is what you can actually turn, and it is always smaller, because the turret and tooling occupy the space above the slide. A machine that “swings 500 mm over the bed” may turn only 280 mm over the cross slide, and the part size that matters is the smaller number. Likewise, for a shaft the real length limit is distance between centres (the tailstock-supported length), not Z-axis travel — read the right row or you will buy a lathe that cannot hold your longest part.

The spindle: power, torque and the numbers that lie

The spindle page carries more misleading rows than any other, because spindle behaviour is a curve, not a number, and the sheet reduces it to a few headline peaks.

Spindle speed. The maximum RPM is a ceiling, and a seductive one. What matters is the usable range for your materials and tool sizes: aluminium and small tools reward a high ceiling, but steel, titanium and large cutters need torque at modest RPM, and a machine with a high RPM ceiling and a weak low end is wrong for a steel shop no matter how fast it sounds. Read the ceiling as the start of the conversation, not the answer.

Spindle taper. BT40/CAT40, BT50/CAT50, HSK and the rest define the toolholder interface the machine accepts and, broadly, its rigidity class — the 40-taper class suits most job-shop work, the 50-taper class is the heavier, stiffer tier. The taper determines which tooling you can use and how easily you can find it, which is why it is a compatibility decision as much as a capability one.

Power — and the peak-versus-continuous trap. A spindle motor can deliver far more than its sustainable output for a short time, because the limit is thermal: the motor generates heat faster than it sheds it, and it can run overloaded only until the windings reach their temperature ceiling. The same physics as a welder’s duty cycle. So a spindle is rated at several levels:

• Continuous (S1) rating — what the motor can deliver indefinitely. This is the number that describes production.
• Time-limited ratings (30 minutes, 15 minutes, 5 minutes) — what it can hold for that duration before overheating.
• Peak rating — the absolute short burst, often two to three times continuous, before protection trips.

The peak figure is the one that sells machines and the one you should ignore for real work: not only because a spindle cannot sustain it, but because the machine structure — bearings, housing, the table beneath — cannot transmit that power to the workpiece without deflecting. Compare on continuous power, and for heavy roughing look at the time-limited rating that matches your actual cycle lengths. Also ask whether the quoted power is measured at the motor or at the spindle: motor-side figures look better and lose a fifth or more to drivetrain losses by the time they reach the cut.

The torque curve. Power gets the spindle spinning; torque keeps it turning when the material pushes back. The row that matters is not a peak number but the shape of the torque-versus-RPM curve: does the machine deliver high torque at the low RPMs where steel and titanium roughing actually happen, or does torque collapse below a mid-range speed, leaving it gutless for big drills and face mills at low RPM? A flat curve across the speeds you cut is worth more than a tall peak you will never reach. Many builders publish the curve; if a sheet gives only numbers, ask for the chart.

Motion: rapids, acceleration and the time you never see

Rapid traverse (the repositioning speed between cuts) is a headline number, and the least honest one on the page. The figure describes top speed, but a machine rarely operates at top speed: it must accelerate to reach it, and with short moves — which is most machining — the axis spends its time accelerating and decelerating and never hits the listed rate. The acceleration/deceleration rates, which the sheet usually buries, decide how fast small moves actually complete. Rapids matter most when moves are long and parts are large; for small-part work, comparing two machines’ rapid numbers is comparing numbers that neither machine will reach. The same logic applies to the adjacent time numbers that do matter for cycle time: tool change time and, on a lathe, turret indexing time. When two machines otherwise match, the machine that changes tools and chips faster wins the cycle.

Accuracy: the rows everyone quotes and few understand

The accuracy block is where the vocabulary matters most, because three different numbers are routinely conflated:

• Resolution is the smallest increment the control can command — 0.001 mm or 0.0001 mm, say. It is a counting number, not a quality number. A fine resolution does not mean a fine machine; the servo and feedback decide how well the machine actually lands where it is told.
• Positioning accuracy is how close the machine gets to where it is commanded, on average and at worst, across the travel. It captures systematic errors — backlash, guideway error, scale error.
• Repeatability is how closely the machine returns to the same position run after run. It is the number that actually governs batch consistency: a machine that always returns to the same spot makes identical parts even if that spot is a few microns from nominal; accuracy can be compensated or probed out, repeatability cannot.

The standard that gives these numbers meaning is ISO 230-2 (with the regional analogues ASME B5.54 and JIS B 6336): the axis is moved to points along its travel, commanded position is compared with measured position by a laser interferometer, the test is repeated many times in both directions, and the statistics yield the quoted accuracy, repeatability and backlash figures. Three things follow. First, the measurement is of the empty machine — no cutting load, no part — so real parts will show more scatter than the sheet. Second, the test is thermal: the machine’s accuracy is stated for a defined temperature state, and thermal drift is the largest source of machining error in long runs — a cold machine and a warm machine are different machines. Ask what the builder does about heat (spindle cooling, structural compensation, refrigerated coolant) and demand the thermal story, not just the warm-room number. Third, ask which standard was used and whether the quoted figure is positioning accuracy of a straight axis or something narrower; a figure without a standard is a figure without meaning.

Two related rows deserve attention on the same page. Feedback type — whether position comes from a motor encoder (semi-closed loop) or from a linear scale mounted on the structure (full-closed loop) — changes how well the machine corrects for screw errors and thermal growth; full-closed feedback is one of the real (and real-cost) upgrades. And contouring accuracy — how well the machine holds a circle or arc — is tested with a ballbar and tells you about the combined behaviour of the axes together, which no single-axis accuracy figure captures. If a builder offers a ballbar or circularity report, read it; it shows the machine as a system, which is how it actually cuts.

The silent specifications: structure and rigidity

The spec sheet’s most important quality rarely appears as a row at all. Rigidity — the machine’s resistance to deflecting under cutting force — is what decides whether the accuracy on the page survives contact with a real part, and it is carried by the things the sheet mentions in passing: machine weight (a heavier machine is usually a stiffer machine — more iron in the castings, more mass to absorb vibration), casting and ribbing quality, and the guideway type (box ways, heavier and more damped for tough cutting, versus linear guideways, faster and lighter). Weight is a crude proxy and an honest one: all else equal, the heavier machine deflects less, and deflection is where accuracy goes to die. When two machines quote identical accuracy, the heavier one with the sturdier ways is very often the better machine — the sheet’s numbers do not show it, but the scale does.

Tooling, workholding and the buried rows

The last block of the sheet decides how usable the machine is day to day. The automatic tool changer capacity and tool change time tell you how many operations a part can run without human intervention and how much of the cycle is spent swapping — capacity matters for unattended running and complex parts (our lights-out guide explains why). On a lathe, the turret station count and whether it offers live (driven) tooling decide whether the machine can mill, drill and tap without a second setup. Watch the maximum tool dimensions and weight (a magazine rated for small tools cannot carry the big cutter your largest part needs), and the workholding interface — table T-slots, pallet or chuck size, and clamping force.

Then read the rows nobody markets: coolant pressure (high-pressure through-spindle coolant is a different machine from a flood-coolant one when it comes to deep holes and difficult alloys), chip handling (conveyor, filtration — the difference between running unattended and shovelling), minimum input increment, air and power requirements, and the machine’s footprint and foundation needs. These buried rows routinely decide whether a machine that looks perfect on paper fits your floor, your utility supply and your workflow. The marketing rows sell the machine; the buried rows determine whether you can run it.

The method: comparing two sheets honestly

When two datasheets are side by side, strip them to the same units and the same definitions before comparing anything:

1. Normalise the rows. Make sure both accuracy figures use the same standard and the same test; both power figures are continuous (or both clearly labelled as their time-limited rating); both power figures are measured at the same point (motor or spindle). A comparison between a peak number and a continuous number is not a comparison.
2. Read the envelope for your largest part, not the sheet’s maximum. Travel, swing-over-carriage or distance-between-centres (whichever is your real limit), nose-to-table, table load — each against the actual worst case from your part portfolio.
3. Read the spindle for your hardest material. Continuous power, the torque curve at the speeds you cut, taper class — against the materials you actually run, not the ones you dream about.
4. Read the accuracy for your tightest tolerance. Which standard, what thermal story, and repeatability ahead of accuracy for batch work. The machine-selection guide gives the four-spec filter this method feeds into.
5. Compare the silents. Weight, guideway type, casting quality, feedback type. The two machines with identical headline numbers are usually separated here.
6. Ask for the reports behind the claims. Accuracy and thermal test reports, a ballbar report, and — above all — a test part cut to your tolerance. The selection guide’s verification step is the final row of every spec sheet worth reading.

Frequently asked questions

What is the difference between accuracy, repeatability and resolution? Resolution is the smallest position increment the control can command — a counting figure, not a quality one. Positioning accuracy (measured per ISO 230-2) is how close the machine gets to where it is commanded, averaged and at worst. Repeatability is how closely it returns to the same position run after run. For production consistency, repeatability is the number that matters most: a machine that always lands in the same place makes identical parts even if that place is a few microns off nominal, and accuracy can be probed or compensated but repeatability cannot.

What does “continuous” versus “peak” spindle power mean? A spindle motor can briefly deliver far more than its sustainable output because the real limit is heat: overload it and the windings heat faster than they shed it. Continuous (S1) power is what it can deliver indefinitely; time-limited ratings (30/15/5 minutes) are what it can hold until the temperature ceiling; peak is the short burst before protection trips. Compare on continuous power, and remember the machine structure cannot transmit the peak figure to the workpiece anyway.

Is machine weight really a specification? It is the best crude proxy for rigidity there is, and rigidity is what decides whether the quoted accuracy survives real cutting load. A heavier machine is usually a stiffer machine — more iron in the castings, more mass damping vibration. When two machines quote identical accuracy, all else equal the heavier one is likely the better machine.

What is the difference between swing over bed and swing over cross slide on a lathe? Swing over bed is the largest diameter the machine can physically clear — the marketing figure. Swing over the carriage (cross slide) is the largest diameter you can actually turn, because the turret and tooling occupy the space above the slide, and it is always smaller. Size your parts against the smaller number, and against distance between centres for long shafts, not Z travel.

How do I compare accuracy figures from two different builders? First check that they used the same standard (ISO 230-2, ASME B5.54, or JIS B 6336) and the same test method — a figure without a stated standard is not comparable. Then remember the figure is for an empty, idle, temperature-controlled machine: ask about the thermal story (how the machine handles heat in long runs) and about repeatability as well as accuracy. Finally, a test part cut to your tolerance is the only number that settles it.

Bottom line

A CNC machine spec sheet is a vocabulary, not a verdict. Learn what each row measures and how it is measured, and the marketing loses its hiding places: advertised maxima turn out to be peaks the structure cannot sustain, “accuracy” turns out to be a claim about an empty machine in a controlled room, and the real differentiators — weight, rigidity, feedback type, thermal behaviour, the torque curve where you actually cut — sit in the rows nobody headlines. Read the family first, the envelope and spindle against your parts, the accuracy against your tolerances with repeatability ahead of accuracy, and the buried rows against your floor and workflow. Then ask for the reports and the test part, because the sheet describes the machine’s best day — the test cut describes the day you will actually work.

This guide is part of the CNC Media guides library — the datasheet-reading reference of the buying topic, deliberately free of prices and of any single builder’s model to promote.