NC and CNC machines explained
This page covers how a machine tool turns numbers into metal removal, where the older hard-wired NC ends and the computer-controlled version begins, and what that difference means for the tolerances and part shapes you can actually order. It is written for design engineers and buyers who need to judge a process, not operate a console.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
What NC and CNC machines explained means at the machine
Every NC and CNC machine does one job: convert a set of coordinates into relative motion between a cutting tool and a workpiece. The coordinates say where the tool tip should be. The machine has to decide how fast to get there, in what order, and with what spindle speed. That decision is the whole difference between the two generations.
On a mechanical NC machine, those coordinates live on a punched tape or a set of cams. The tape is read block by block, and the machine follows a fixed sequence. Change one dimension and you cut a new tape. There is no memory, no editing, no simulation. The machine does not know what part it is making; it only knows the next set of pulses.
A CNC machine stores the same information as a digital file. The controller holds the program in memory, runs look-ahead, and adjusts feed and acceleration while the cut is happening. You can edit one line and re-run. That single change is why complex geometry became practical, and why a shop can hold ±0.005 mm on a production run instead of only on a one-off.
So when people ask about NC and CNC machines explained in one sentence, the answer is this: NC executes a frozen instruction stream, CNC executes a live one. Everything else, from tool compensation to five-axis motion, follows from that.
- 1NCHard-wired or tape-fed, fixed sequence, edit means re-cut the medium.
- 2CNCProgram stored digitally, editable, with real-time compensation.
- 3Shared groundBoth use the same axis convention and the same G-code family.
How a controller turns a file into a cut
The chain starts in CAD. A solid model or a 2D drawing defines the geometry. CAM software then decides which tool removes which volume, in what order, and writes that plan as G-code: rapid moves, feed moves, arcs, tool changes, spindle commands. The post-processor matters here, because it translates generic CAM output into the dialect your controller expects.
The controller reads that code and interpolates. If the program says move from X0 Y0 to X50 Y30, the controller calculates the intermediate points, checks them against the acceleration limits of each axis, and outputs position commands to the servo drives. A good controller looks ahead several blocks so it can slow into a corner before it overshoots it. That look-ahead is why a modern machine can run a 3D contour at a constant chip load instead of stuttering at every node.
Feedback closes the loop. Rotary encoders or linear scales report actual position back to the drive, which corrects the error in real time. On a machine with linear scales, the measurement is taken at the table, so it catches ballscrew growth from heat. On a machine with only motor encoders, thermal drift stays invisible to the control.
Then the physical side takes over. Rigidity, spindle runout, tool holder balance, and thermal stability set the floor on what the control can achieve. A perfect program on a flexing machine still cuts a tapered bore.
- 1CADDefines nominal geometry and tolerances.
- 2CAMChooses toolpaths, feeds, and stepover.
- 3Post-processorMatches output to the specific controller dialect.
- 4Servo loopCompares commanded and measured position thousands of times per second.
Three, four, and five axis: what each one buys you
A three-axis machine moves the tool in X, Y, and Z only. The part sits still or indexes between setups. This is the workhorse for plates, brackets, housings, and any feature you can reach from one direction. It is also the cheapest way to hold tight tolerance because there are fewer moving elements to stack error.
A four-axis machine adds rotation about one axis, usually A or B. That lets you machine around a cylindrical part, cut slots at several angular positions, or reach four faces without re-fixturing. If your part is a shaft, a manifold, or a connector body with features on its sides, four-axis removes a setup and removes the position error that comes with it.
Five-axis adds a second rotary axis, so the tool can approach the part from almost any direction. The real gain is not just access. It is that a shorter, stiffer tool can reach deep features, and that you can tilt the tool to control where the cutting edge contacts the surface. That matters for impellers, turbine blades, and any contoured surface where a ball nose tool would otherwise rub at its center.
The trade-off is real. Five-axis machines cost more per hour, need more careful programming, and demand rigid fixturing because the part moves. On a simple prismatic part, five-axis adds cost without adding capability.
- 1Three-axisBest for prismatic parts reachable from one or two directions.
- 2Four-axisBest for cylindrical or multi-face parts that would need re-fixturing.
- 3Five-axisBest for contoured surfaces, deep cavities, and single-setup complex geometry.
Where these machines reach their limit
No controller can fix a bad setup. If a part is held in a vise with 0.05 mm of lift on one corner, the machine will faithfully cut that error into every face. The first check on any tight-tolerance job is the fixture, not the program.
Tool length to diameter ratio sets another ceiling. A 6 mm end mill hanging 60 mm out of the holder will deflect under cutting load, and no amount of look-ahead compensates for it. For deep pockets, the answer is usually a larger tool with a relieved shank, or a five-axis approach that keeps the tool short.
Thermal drift is the quiet one. A spindle running at 15,000 rpm for two hours grows, and so does the ballscrew. On a machine with linear scales and a temperature-controlled shop, the effect is small. In a shop that swings 8 °C between morning and afternoon, it is not. This is why long production runs are often scheduled with a warm-up cycle and periodic in-process checks.
Material matters too. Aluminum 6061 and 7075 cut freely and forgive a lot. Titanium Ti-6Al-4V and Inconel generate heat at the cutting edge, work-harden if the feed is too light, and will destroy a tool in minutes if the parameters are wrong. On those materials, the process window is narrow and the machine's rigidity matters more than its control.
- 1Fixture errorShows up as a dimensional error the control cannot see.
- 2Tool deflectionScales roughly with the cube of the length-to-diameter ratio.
- 3Thermal growthCaught by linear scales, missed by motor encoders alone.
What this means when you place an order
For a buyer, the practical question is not which generation is better. It is which machine can hold your tolerance on your geometry at a cost that makes sense. A flat plate with six holes does not need five-axis. A contoured impeller with 0.02 mm blade thickness does not survive on three-axis without a lot of hand work.
Send the 3D model and the drawing with tolerances, not just the model. The tolerances tell the shop which features are critical and which are free. That determines the machine, the fixture, and the inspection plan. A model without tolerances forces the shop to guess, and guessing usually means quoting the tightest case.
Ask about the inspection method for the critical features. A caliper reading is not the same as a CMM report, and a first-article report is not the same as 100% inspection. For medical and automotive work, the documentation is often as important as the part.
At GreatLight, we run 127 high-precision CNC machines across three plants in Dongguan and Singapore, including 16 simultaneous five-axis machining centers and 16 mill-turn centers. Maximum processing size is 4,000 mm, and we hold ±0.005 mm with 100% inspection before shipment. We quote and return a free DFM analysis within 12 hours, and production can start within 24 hours.
Uploads are secure and confidential, and an NDA is available on request.
- 1Send tolerancesNot just the model. It sets the machine and the inspection plan.
- 2Name the critical featuresSo the shop knows where to spend time and where not to.
- 3Ask for the inspection methodCaliper, CMM, or full report. They are not equivalent.
NC versus CNC: where the two actually diverge
The differences that change what you can order
| Aspect | NC machine | CNC machine |
|---|---|---|
| Program storage | Punched tape or cams | Digital file in controller memory |
| Editing a dimension | Cut and re-punch the tape | Change one line and re-run |
| Tool compensation | Manual offset by the operator | Cutter comp and wear offsets in the control |
| Geometry ceiling | Straight lines, simple arcs | Free-form surfaces, 3D contours |
| Repeatability | Drifts with tape wear and setup | Holds ±0.005 mm across a run |
| Setup change | Hours, often a new tape | Minutes, load the next program |
| Typical use today | Legacy lathes, training, hobby | Production, prototype, five-axis work |
The short version
If your part is prismatic and reachable from one or two directions, a three-axis machine is the cheaper, more rigid choice. If it has contoured surfaces, deep cavities, or features on five sides that must stay in one setup, five-axis pays for itself in fewer fixtures and tighter true position.
Questions engineers ask next
Can a CNC machine run without a CAM program?
Yes. Simple features can be programmed at the control with manual G-code, or with canned cycles for drilling and tapping. This is common for one-off parts, repairs, and quick modifications.
For anything with a 3D surface, CAM is the practical route. Hand-writing a toolpath for a contoured surface is slow and error-prone, and the control has no way to verify it before the cut.
Does five-axis always give better tolerance?
No. Five-axis improves access and can reduce the number of setups, which removes stacking error from re-fixturing. On a part that already fits in one three-axis setup, adding rotary axes adds error sources rather than removing them.
The tolerance gain shows up on parts that would otherwise need three or four setups, or on deep features where a short, tilted tool is stiffer than a long, straight one.
What surface finish can I expect as-machined?
As-machined finishes typically land around Ra 1.6–3.2 μm. With careful parameters and a finishing pass, Ra 0.8–1.6 μm is achievable on many materials.
Fine finishes down to Ra 0.2–0.8 μm are possible on specific features with the right tool and a light finishing cut. If the finish is functional, say so on the drawing, because it changes the toolpath and the cycle time.
Why does my part move during machining?
Usually it is the fixture, not the program. Thin walls deflect under clamping pressure, and a part held only on one face can lift as the tool pushes sideways.
The fix is often a support under the unsupported area, a softer clamping method, or a roughing pass that leaves more material for the finishing pass to remove symmetrically.
How do I know if my tolerances are realistic?
Check the tolerance against the feature size and the material. A ±0.005 mm callout on a 500 mm aluminum part is harder than the same callout on a 30 mm steel pin, because thermal and deflection effects scale with length.
A DFM review will flag callouts that drive cost without adding function. We return that review free with the quote, usually within 12 hours.
Does the controller brand affect the part I get?
Mostly it affects the shop, not the part. Different controllers use different dialects and different look-ahead behavior, so a program may need a different post-processor.
What reaches you is the same geometry either way, as long as the machine is mechanically sound and the inspection plan matches the tolerances on the drawing.
Send a model, get a straight answer
Upload your CAD file and tolerances. We reply with a quote and a free DFM analysis within 12 hours, and production can start within 24 hours.
12-hour quote100% inspectionNDA on request