Choosing between DMLS (direct metal laser sintering) and CNC machining for a metal part is not a whole-part decision — it is a feature-by-feature one. At Layer X, the honest answer to "which process is cheaper?" is that internal cooling channels, buried undercuts and topology-optimised lattices belong on DMLS, while flat sealing faces, large simple prisms and high-volume repeat parts belong on a CNC mill. Most real components carry both kinds of feature, which is why our standard DMLS tolerance is ±0.1mm with critical surfaces then post-machined below ±0.05mm. This guide gives you the decision matrix we use internally, so you can route each feature to the right process before you request a quote.
Why "pros and cons" lists mislead you
Generic comparisons tell you DMLS is good for complexity and CNC is good for tolerance. True, but useless at the point of decision — because a single bracket can have a conformal cooling channel that only DMLS can build and a bearing bore that only precision machining can hold. The cost and the tolerance both live at the feature level, not the part level. So instead of asking "DMLS or CNC for this part?", ask "DMLS or CNC for this feature?" and let the geometry decide. The three variables that settle almost every case are tool access (can a cutter physically reach the surface?), internal geometry (are there enclosed voids?) and batch size (how many identical parts?).
The feature-to-process decision matrix
This is the table we run every incoming metal enquiry against. Read down the feature column, and the recommended route follows from the geometry — not from a preference for either machine.
| Part feature | Route to | Why |
|---|---|---|
| Internal cooling / fluid channels (conformal) | DMLS | Enclosed curved voids cannot be reached by any cutter; DMLS builds them layer by layer |
| Buried undercuts with no line-of-sight tool access | DMLS | A mill needs to reach the surface; if it cannot, the feature is unmachinable |
| External undercuts, reachable with a 5-axis approach | CNC | Tool access exists, so subtractive holds tighter tolerance more cheaply |
| Flat sealing faces, bearing bores, threaded holes | CNC (or DMLS + post-machining) | Sub-0.05mm flatness and surface finish are a machining strength |
| Topology-optimised lattices / organic thin walls | DMLS | Freeform mass reduction that would take a mill days, or is simply uncuttable |
| Large, simple prismatic bodies (blocks, plates) | CNC | Removing bulk material from stock is fast and cheap; building it in powder is slow |
| Multi-part assemblies consolidated into one body | DMLS | Fewer joints, no fasteners; the geometry is impossible to machine as one piece |
| Identical parts, 500+ units, simple geometry | CNC | Per-part machining time falls with fixturing and automation; DMLS build time does not |
Undercuts: the tool-access test
An undercut is any surface a straight-line cutter cannot reach from the outside. The mistake is treating all undercuts the same. Draw a line from the outside of the part to the surface in question: if a rigid tool can travel that line without colliding with the rest of the body, a 3- or 5-axis mill can cut it, and machining will hold a tighter tolerance for less money. If that line is blocked — the classic case being an undercut on the inside of an enclosed cavity — no cutter exists that can reach it, and the feature is only producible by an additive process like DMLS. So an undercut does not automatically mean 3D printing. It means check the tool path first. This is exactly why hybrid tool inserts, which combine buried cooling geometry with precision cavity faces, are so often 55–65% cheaper to produce additively than to machine from solid — the undercut features that explode CNC time add almost nothing to a DMLS build. We cover that economics in depth in our injection tooling service.
Internal channels: the feature DMLS owns outright
Enclosed internal channels are the cleanest case in the whole matrix, because there is no argument to have. Conformal cooling lines that follow the contour of a mould cavity, curved fluid manifolds, gas passages inside a valve body — these are closed, curved voids. Drilling can only produce straight intersecting bores, which force sharp corners, dead volumes and plugged access holes. DMLS grows the channel as part of the solid, in any path the flow analysis asks for. If your part has a true internal channel that is not a straight through-hole, the decision is already made: it goes to DMLS metal 3D printing, and any flat or bored feature elsewhere on the part is cleaned up by post-machining afterwards. This is the standard workflow behind our sub-0.05mm critical-surface tolerance — build the impossible geometry, then machine the faces that need to seal.
Batch size: where the cost curves cross
DMLS build time is roughly fixed per part — the laser has to fuse the same volume of powder whether it is your first part or your five-hundredth. CNC behaves the opposite way: the first part carries all the fixturing and programming setup, and every part after that gets cheaper as that cost amortises. So the two cost curves cross. Below the crossover, DMLS wins on total cost because there is no tooling or fixture to pay off. Above it, CNC wins because per-part machining time keeps falling while DMLS per-part cost stays flat.
| Batch size | Simple geometry | Complex geometry (channels / undercuts / lattice) |
|---|---|---|
| 1–10 parts | Either; CNC often cheaper for plain blocks | DMLS — no tooling to amortise |
| 10–500 parts | CNC as volume rises | DMLS — machining the complexity stays expensive at every quantity |
| 500+ parts | CNC, or bridge tooling into moulding | DMLS for validation, then hybrid tooling for production runs |
The nuance the simple curve misses is that complexity moves the crossover. For a plain prismatic block, CNC overtakes DMLS at low quantities. For a part full of internal channels, machining stays expensive at every quantity, so the crossover moves far to the right — sometimes past the point where you would move to bridge tooling anyway.
Tolerance, honestly stated
Set expectations with the real numbers rather than the marketing ones. Standard as-built DMLS tolerance at Layer X is ±0.1mm across general geometry — good, but not machining-grade. When a feature needs to be tighter, we post-machine that specific surface to below ±0.05mm and verify it on a CMM, with the dimensional report supplied on every order. Precision CNC machining works to general tolerance grades such as those in ISO 2768, and holds close tolerances on individual reachable features directly. Additive tolerances and test methods, in turn, are framed by ISO/ASTM 52900, the base terminology standard for additive manufacturing. The practical takeaway: do not spread a ±0.02mm callout across an entire DMLS part. Put the tight tolerance only on the faces that truly need it, so those become the post-machining or CNC operations, while the freeform bulk stays additive.
| Attribute | DMLS (as built) | DMLS + post-machining | Precision CNC feature |
|---|---|---|---|
| Typical tolerance | ±0.1mm | Below ±0.05mm on the machined face | Close-grade on reachable features |
| Enclosed internal geometry | Yes | Yes | No |
| Best for | Complex, low-to-mid volume | Complex bodies with critical faces | Simple / high-volume, tight faces |
Material and certification realities
Route decisions also touch the alloy. Our DMLS system runs Ti-6Al-4V, 316L and 17-4 PH stainless steels, Inconel 625 and 718, and H13 tool steel — with material certificates, CMM reports and a certificate of conformance on every order. That documentation matters because Layer X is AS9100 Rev D, ISO 9001:2015 and ISO 13485:2016 certified, and works to an overall studio tolerance of ±50µm with a 99.4% first-pass yield across 2,000+ parts shipped. If your part is a large mild-steel or aluminium enclosure rather than a fully dense functional alloy component, the answer may not be either milling or DMLS at all — it may be laser cutting into CNC press-brake bending, produced single-source in the same facility. Match the process to the material and the duty, not to the buzzword.
A five-question routing checklist
- Does any surface sit inside an enclosed void? Yes → DMLS.
- Is there a true internal channel that is not a straight through-hole? Yes → DMLS.
- Can a cutter physically reach every critical face? Yes, and geometry is simple → CNC.
- How many identical parts? Hundreds of a simple part → CNC; tens of a complex part → DMLS.
- Where are the tight tolerances? Isolate them to specific faces and route those to machining or post-machining, not the whole part.
Run a part through those five questions and it usually sorts itself. For the cases that split — a complex body with a couple of precision faces — the winning answer is almost always the hybrid one: build additively, finish the faces that matter. For the wider trade-offs on cost, see our guide to reducing 3D printing costs in India, our tolerances and dimensional accuracy guide, and the DMLS guide for aerospace and defence.
Not sure which way a feature should route? Upload your CAD file for a 24-hour quote and we will mark up, feature by feature, exactly what we would build on DMLS and what we would machine — with the tolerances and pricing spelled out before you commit.