CNC or 3D Printing: Which Path Should a Drawing Take? Five Criteria to Decide

A bracket drawing: why two suppliers quoted a threefold price difference

Last week a drawing for a 200×80 aluminum bracket was sent out for quotation, and the numbers that came back were interesting: one supplier quoted CNC milling at 480 yuan per piece; another quoted 3D printing at 160 yuan per piece. The customer came with both quotes and asked: is the second supplier cutting corners?

Looking at the drawing makes it clear. The part has two Ø8H7 pin holes, one Ra1.6 sealing mounting surface, and the rest are connecting ribs and lightening slots. The drawing only marked "aluminum part," without specifying functional surface requirements or batch size. The CNC supplier worked to the fit surfaces, with two setups, boring plus fine milling, and could not bring the labor hours down; the 3D printing supplier worked to a cosmetic part, one-piece molding, and only sanded the support surfaces during post-processing. Neither supplier was wrong—what was wrong was that the drawing did not clearly state "which areas are functional surfaces" and "how many pieces are needed."

This kind of price difference is not a supplier ripping you off, but a cost structure difference caused by different process routes. The bulk of CNC cost lies in fixturing, tooling, and labor hours; the bulk of 3D printing cost lies in material volume, supports, and post-processing. The same geometry, two routes, can differ in price by two to three times, and delivery quality is also completely different.

So selection cannot start from "which is cheaper," but must be worked backward from "which surfaces must be preserved, how many pieces are needed, and what material." Go through the following five criteria in order, and you can basically lock in which route to take, or whether to take a hybrid route.

Subtractive vs Additive: Understand the Principle Difference First, Then Discuss Selection

For the same bracket drawing, the CNC quote says "blank 6061-T6, 3-axis milling, deburring," while the 3D printing quote says "SLS nylon, depowdering, sandblasting." Both can produce the outer shape, but internal stress, surface texture, and hole-wall finish are completely different. Selection is not about comparing prices, but about comparing physical processes—whether a tool removes material or material is stacked layer by layer. This difference determines all the criteria that follow.

CNC is subtractive: the tool cuts chips from a solid blank, and material removal rate, cutting force, and fixturing rigidity directly determine how deep, how fast, and how thin you can cut. 3D printing is additive: a laser or nozzle stacks material layer by layer, with each layer thickness usually between 0.02–0.2 mm (depending on the process, confirm with the equipment), and interlayer bonding strength is often lower than that of the base material itself. The former is limited by tool interference and clamping, while the latter is limited by support structures and thermal history.

Design Freedom: Undercuts, Internal Channels, and Lattices—Who Can Make Them

CNC cannot make closed internal cavities where the tool cannot reach, nor narrow slots with excessive depth-to-width ratios; 3D printing is almost unconstrained by geometry, and internal channels, lattices, and integrated hinges can all be formed in one piece. But printed parts need supports, and overhang angles are generally recommended to be kept within 45°, otherwise bottom surface roughness will deteriorate noticeably. If the drawing has cross-drilled holes or internal cooling channels, first ask yourself: can CNC drill them from both ends to meet? If yes, use CNC; if not, then consider additive.

Material and Residual Stress: Anisotropy vs Rolling Direction

CNC uses rolled or forged blanks, with dense material and good isotropy, and many metal grades are available (aluminum alloy, stainless steel, titanium, copper, engineering plastics). 3D printed plastic parts are anisotropic: strength in the XY direction is usually higher than in the Z direction, and the interlayer is the weak plane. Metal printing (DMLS) can achieve very high density, but thermal stress easily causes warping, requiring substrate preheating or subsequent heat treatment. If the part bears alternating loads or impact, prioritize CNC or densification treatment after printing.

Note: The tolerances and surface roughness in the table above are common industry references, not a commitment from any factory. What can actually be achieved depends on feature size, fixturing method, toolpath, and inspection tools. When the drawing specifies below ±0.05, first confirm the datum and inspection method, then discuss the process.

Criterion 1: Functional surfaces and tolerance zones determine the process lower limit

The most easily overlooked thing on a drawing is "which surfaces actually do the work." Mating surfaces, seal grooves, sliding guideways, load-bearing holes—the tolerance zones and surface roughness of these surfaces directly set the process lower limit. For 3D printing (FDM/SLS/MJF), dimensional deviation in the build direction is commonly ±0.3~0.5 mm/100 mm, and layer stepping Ra is often 12~25 μm; while CNC finish milling of mating surfaces can conventionally be controlled within ±0.02~0.05 mm in the industry, and precision grinding can be even tighter. If the drawing specifies only one overall tolerance without distinguishing functional surfaces, both quoting and process selection will go off track.

The method of judgment is simple: divide the surfaces on the drawing into three categories—mating/sealing/sliding/load-bearing must go through CNC; appearance surfaces can be printed and then sanded; non-critical support surfaces can simply be printed. A common mistake is to mark the entire part as ±0.05, resulting in 3D printing being unable to achieve it while CNC becomes absurdly expensive. The correct approach is to tighten tolerances only on functional surfaces and release the rest according to general tolerances (such as GB/T 1804-m).

Correspondence between tolerance zones and inspection methods

Before specifying a tolerance, first ask: what will be used to measure this tolerance? A caliper has a resolution of 0.02 mm, so measuring ±0.05 is already marginal; below ±0.02 requires a micrometer, CMM, or optical scanning. If the factory does not have the corresponding inspection tools, specifying it is useless. A more hidden pitfall is the datum: if the drawing specifies position tolerance but does not specify the datum plane, CNC fixturing can only guess. This is even more true for printed parts—interlayer bonding surfaces are naturally anisotropic, and choosing the wrong datum plane can double hole position deviation.

Rules for using this table: first see which row the functional surface tolerance falls into, then decide the process route. If it falls below ±0.02, prioritize CNC; if the part geometry is so complex that CNC fixturing is difficult, then evaluate the hybrid route of "3D printed blank + CNC finishing of functional surfaces." Note: The values in the table are common industry references, not committed values from any factory; before quoting, be sure to have the supplier confirm whether their equipment and inspection tools can cover your tolerance zone.

Final reminder: if the drawing shows "mating surface Ra 0.8" but no machining symbol, or "hole position ±0.05" but no datum, do not rush to place an order—this kind of missing information is a major source of later disputes. Circle the functional surfaces, mark the tolerance zones clearly, and only then will process selection have a basis.

Criterion 2: Batch size and geometric complexity determine the cost curve

The unit price curve is not linear. For the same drawing, the cost structure for 1 piece and 500 pieces is completely different: the cost of 3D printing is almost entirely in material and machine time, with fixturing and programming close to zero; in CNC cost, "fixed overheads" such as programming, fixtures, tool setting, and first-article inspection account for the majority, while per-piece material cost is secondary. Therefore, when judging the choice, first ask about batch size, then ask about geometry.

Batch-size watershed: printing is advantageous within a few dozen pieces; above a few hundred pieces, simple-geometry CNC overtakes it

A common industry situation is: 1–50 pieces, 3D printing (especially SLS/MJF plastic parts or DMLS metal parts) eliminates fixtures and tooling, making unit cost flatter; 50–200 pieces is a transition zone, depending on part geometry and material—for simple rotational parts and plate-type parts, CNC can run with soft jaws or a vacuum chuck, and unit price begins to fall below printing; above 200 pieces, CNC for simple geometries is usually clearly more economical because fixed overheads are diluted, while printing machine time and material costs hardly decrease. This is not a hard boundary; the more complex the drawing and the harder the material is to print, the more the watershed shifts toward printing.

Do not ignore the impact of geometric complexity on the shape of the cost curve. CNC cost rises quickly with the number of features, undercuts, deep cavities, and thin walls—each additional surface requires another fixturing setup or tool change; five-axis can reduce fixturing but machine time cost is also high. 3D printing cost is almost insensitive to complexity: internal channels, lattices, integrated hinges, conformal cooling channels—these are either impossible on CNC or require splitting, welding, and then machining, while in additive they are just a few more slicing layers. So when a part contains features that "only additive can make," the batch-size watershed fails, and you should go directly to printing or a hybrid route.

A pitfall that's easy to fall into: equating "large batch size" directly with "CNC is cheaper." If the part is thin-walled, prone to deformation, and requires multiple setups, CNC's scrap rate and auxiliary labor hours may eat up the batch advantage. Conversely, if the printed part requires extensive post-processing (support removal, sanding, infiltration, heat treatment), the unit cost will also climb. When calculating cost, ask the supplier: does the quote include post-processing, first-article inspection, and fixturing costs?

If the geometric complexity is already high enough that CNC requires splitting the part, or the batch size falls in the transition zone and you're unsure, you can submit the drawing to both processes for comparison. In the AI chat quote, clearly state the batch size, material, and key tolerances—this usually gets you comparable decision-making information faster than just sending a 3D file.

Criterion three: material and post-processing determine whether it can be used

Even if the first two criteria are met, the drawing may not be manufacturable—material grades and post-processing often only surface problems after quoting. CNC can handle grades like 6061, 7075, 304, 316L, POM, PEEK, etc., with consistent performance across batches of the same grade; 3D-printed plastics (PLA, ABS, PETG, nylon, resin) exhibit anisotropy: interlayer strength is often lower than in the XY direction, and nylon parts drift in dimensions after absorbing moisture. Metal printing (DMLS) can produce 316L, Ti64, AlSi10Mg, but density, residual stress, and support removal all affect final performance. Before choosing a material, ask first: how does this part fail? If it's interlayer delamination, the printed part needs a different orientation or should switch to CNC.

Achievability differences between metal and plastic

CNC metal parts perform close to forgings, suitable for load-bearing, sealing, and sliding fits; 3D-printed plastic parts are better suited for enclosures, fixtures, and non-load-bearing structures. Metal printing can handle complex channels and lightweight lattices, but surface roughness, internal stress, and thread strength are usually inferior to CNC. If the part requires anodizing, electroplating, or heat treatment, confirm the process window first—printed parts may deform during heat treatment, requiring allowance or post-machining.

Post-processing is not optional; it is part of the function

The steps left by support removal, the uniformity of sandblasting, and the duration of post-curing all change dimensions and appearance. If the drawing only states "no surface defects," the supplier can only rely on experience. It is recommended to specify on the drawing: the Ra requirement for functional surfaces, whether support marks are allowed, the heat treatment condition (such as T6), and whether coloring or electroplating is required. When unsure, first run a small-batch trial to verify, then define the post-processing specification.

After materials and post-processing are confirmed, return to the quoting stage. If the drawing specifies a particular grade and heat treatment condition, AI chat quoting can match the process faster; if you are still at the material selection stage, you can first check the materials library to confirm feasibility.

Hybrid route: how to plan 3D-printed blanks + CNC finishing

For some parts, a single process is unsatisfactory on both ends. For example, an injection mold insert with internal conformal cooling channels: with full CNC milling, the channels must be split into multiple pieces and welded together, resulting in many sealing surfaces and long lead times; with full metal printing, the channels are formed in one piece, but the roughness and dimensional accuracy of mating surfaces and sealing surfaces cannot meet assembly requirements. In this case, a hybrid route—printing the blank and CNC-finishing the functional surfaces—is often more cost-effective than forcing a single process.

Process sequence: print first, then finish machine

Small case: a mold insert with channels

The part envelope is 80×60×40 mm, material 1.2709 mold steel, internal conformal channel diameter 6 mm, with the channel 4 mm from the nearest surface. The mating surface requires flatness of 0.02 mm, and the sealing groove is 3 mm wide with a tolerance of ±0.02 mm.

Plan: metal print the entire blank, forming the outer shape and channels in one piece; leave 0.5 mm allowance per side on the mating surface, sealing groove, and ejector pin holes; perform stress-relief annealing after printing; on the CNC, first mill the bottom face and two datum side faces, then flip and finish-mill the mating surface, finish-mill the sealing groove, and drill the ejector pin holes. The channels are not machined and retain the as-printed surface—the roughness of the channel inner wall affects cooling efficiency, but usually does not require a mirror finish, so the as-printed state is acceptable; if the channel requires Ra 1.6 or lower, abrasive flow polishing or electrochemical polishing must be added, which is post-processing and not something CNC can solve.

The key judgment here is: which surfaces must be CNC-machined, and which surfaces can remain as-printed. Mating surfaces, sealing surfaces, sliding surfaces, and load-bearing surfaces must be finish-machined; features that tools cannot reach, such as channels, lattices, and internal cavities, are final in the as-printed state. How much allowance to leave depends on the dimensional deviation and surface roughness of the printing process—too little, and finishing cannot remove layer lines or warpage; too much, and cutting time increases, and thin-walled parts may deform due to cutting forces. A common industry practice is 0.3–0.8 mm per side, but the final value must be confirmed according to the printing equipment capability and part stiffness.

The hybrid route actually has boundaries for suitable scenarios: internal features are too complex for CNC to machine, while the exterior also has a small number of high-precision functional surfaces. If the internal features can also be made by CNC (such as straight holes or simple cavities), or if the external functional surface precision requirements are not high (the as-printed state can satisfy them), then the hybrid route is just unnecessary stacking of operations. The judgment order is: first check whether the functional surface tolerances can be met in the as-printed state; if yes, use a single process; if not, then check whether the internal features can be made by CNC; if yes, use single-process CNC; only when both ends are blocked should the hybrid route be considered. This order can avoid spending money on two extra operations just to be "advanced."

Failure modes and inspection: what happens on site after choosing the wrong process

Choosing the wrong process will not be exposed immediately on the quotation sheet, but on the assembly line. The following is organized into four columns: failure phenomenon, common root cause, inspection method, and procurement/engineering action, directly corresponding to the tolerances, datums, and batch size on the drawing.

The correspondence between inspection methods and drawing tolerances

Not every dimension is worth putting on a CMM. Calipers can measure to the ±0.1 mm level, micrometers to ±0.01 mm, and only CMMs and 3D scanning can reliably cover mating dimensions within ±0.02 mm. If the drawing specifies tight tolerances but has no datum or gauge definitions, the supplier can only measure according to their own interpretation, and the result is that "pass" and "fail" each tell their own story. A common industry practice is: mark datums A/B/C for critical mating dimensions and specify the inspection method; non-functional dimensions only need general tolerances.

Procurement implications: incoming inspection and supplier capability confirmation

Before choosing a process, first confirm whether the supplier can inspect it. For printed parts, ask clearly whether they have a post-curing oven and can provide a scan report; for CNC parts, ask clearly whether they have a CMM and can do a full-dimensional first article. For incoming inspection, don't just measure the outer shape—list mating holes, mounting surfaces, and threaded pilot holes as mandatory inspection items. When the batch exceeds a few dozen pieces, require the supplier to provide a first article report and SPC data for critical dimensions; otherwise, later assembly interference will leave you to rework it yourself.

If a failure has already occurred, don't rush to change the process. Use 3D scanning to compare the actual part with the CAD to determine whether it is a dimensional deviation or a geometric deviation; then check the clamping records and printing parameters. In most cases of "the process doesn't work," the root cause is that the datums were not unified or a post-processing step was missed. Getting the inspection data back is more useful than requesting a new quote.

Frequently Asked Questions (FAQ)

The following questions are the ones that appear most frequently in drawing reviews and quote exchanges. The answers all come with preconditions—check them against your own part before copying the conclusions.