Climb and conventional milling are the two ways a side-milling cut can meet the material, and in G-code they are chosen by geometry, not by a dedicated code: with the spindle running forward (M03) and standard right-hand tooling, the rule set is fixed. Cut an outside profile traveling clockwise around the part, or an inside pocket traveling counterclockwise, and you are climb milling; reverse either direction and you are conventional milling. In cutter-compensation terms, G41 puts standard tooling in climb, G42 in conventional. Climb is the modern default on CNC machines, better finish, longer tool life, chips thrown clear of the cut, while conventional keeps a place in specific situations: worn machines with backlash, abrasive skins on castings and hot-rolled stock, and moments when the grab of a climb entry would deflect a flimsy setup. Knowing which you are commanding, from the code alone, is a small skill that explains a long list of finish and tool-life mysteries.
What the two modes physically are
The difference lives in how each tooth meets the material. In climb milling the tooth enters at full chip thickness and exits thin: it bites immediately, forms its chip, and leaves cleanly. In conventional milling the tooth enters at zero thickness and thickens as it sweeps: for the first fraction of its arc it is not so much cutting as rubbing, sliding over the surface until enough material accumulates ahead of the edge to shear. The milling cutter overview covers the geometry; the consequences follow directly from it.
That rubbing entry is why conventional milling runs hotter, wears tools faster on most materials, and work-hardens alloys prone to it, stainless steels being the famous case, where each rubbing pass hardens the very surface the next tooth must enter. Climb’s clean entry produces lower cutting forces on the finished surface, chips evacuated behind the cutter rather than dragged through the cut ahead of it, and the measurably better finish that makes it the default recommendation everywhere rigidity allows. Climb’s cost is directional force: it pulls the cutter into the work along the feed direction, which is harmless on a machine that can resist the pull and problematic on one that cannot.
The G-code mapping: direction decides everything
No G or M code says climb; the mode falls out of three things the program does state: spindle direction, path direction, and which side of the material the cutter runs. With M03 forward rotation and ordinary right-hand tools, the complete practical map is:
| Cut | Travel direction (viewed from above) | Comp side | Mode |
|---|---|---|---|
| Outside profile | Clockwise around the part | G41 (left) | Climb |
| Outside profile | Counterclockwise around the part | G42 (right) | Conventional |
| Inside pocket wall | Counterclockwise around the pocket | G41 (left) | Climb |
| Inside pocket wall | Clockwise around the pocket | G42 (right) | Conventional |
| Open slot / side pass | Material on the right of travel | G41 (left) | Climb |
| Open slot / side pass | Material on the left of travel | G42 (right) | Conventional |
The compensation column is the compact version: with forward spindle and right-hand tooling, G41 means climb and G42 means conventional, which is why experienced readers judge a program’s milling mode from its comp calls at a glance. Everything inverts together if something upstream inverts: run the spindle in M04 reverse with left-hand tooling and the whole table flips, which is rare on mills but a real configuration on some setups. The reliable habit is not to memorize six rows but to reconstruct any row in five seconds: face the direction of travel, ask which side the material is on, and remember that material on the right of travel is climb for a forward-running spindle.
CAM systems make the same choice through a checkbox, climb or conventional per operation, and then emit the path direction and comp side that produce it. That buried checkbox is worth auditing when a toolpath misbehaves, because a copied operation or a template default can silently carry the wrong mode, and the symptom appears at the machine as finish or tool life rather than as anything visibly wrong in the code, part of the broader skill of reading programs to find problems.
When climb is right, and when it bites
On a rigid CNC machine with healthy ballscrews, climb is the default for both roughing and finishing, and the reasons compound: tool life from the clean entry, finish from low forces at the finished wall, chip evacuation away from the cut, lower heat into the part. Modern tooling assumes it; feeds-and-speeds recommendations, built on the chip-load arithmetic of standard speeds and feeds, are calibrated to climb engagement unless stated otherwise.
Climb bites in three circumstances worth knowing cold. Backlash: climb’s pull-into-the-cut reverses load on the drive mechanism, and on a machine with lash, a worn manual knee mill, a loose hobby machine, a neglected router, the table can be snatched forward by the cutter, taking a sudden oversized bite that at best gouges and at worst breaks the tool. This is the historical reason conventional milling was the manual-machinist’s rule, and it remains true wherever backlash lives; the same snatch mechanic shows up in why an endmill breaks on G01 stories on flexible machines. Hard skins: castings, forgings, flame-cut plate, and hot-rolled bar wear an abrasive crust, and a climb entry drives the tooth down through that crust at full thickness each engagement, while a conventional pass slices up from beneath it, protecting the edge. Deflection-critical passes: climb’s pull can draw a slender tool or a thin wall into the cut; some finishing situations on flexible parts run a conventional spring pass precisely because its forces push away rather than pull in.
The honest summary is asymmetric: on CNC equipment in good condition, exceptions to climb are deliberate and occasional; on backlash-prone machines, conventional is not a preference but a safety rule.
A concrete pair of program fragments makes the mapping tactile. The same square boss, finished 0.5 inch deep with a half-inch end mill, first as a climb pass:
(CLIMB: outside profile, clockwise, G41)
G00 X-0.6 Y-0.6
G01 Z-0.5 F30.0
G41 D1 G01 X0 Y-0.25 F12.0 (lead in, comp left)
G01 X2.0 (along bottom edge, part on right)
G01 Y2.25
G01 X0?That question mark is deliberate, because writing the fragment from memory is the test: traveling the boss clockwise, each successive edge keeps the part on the tool’s right, and the coordinates follow. The conventional version of the same boss reverses the direction of travel, counterclockwise, and swaps the comp call to G42; every coordinate pair simply appears in the opposite order. Two programs, identical geometry, opposite modes, and nothing but direction and comp side distinguishing them. Running both on a piece of scrap and comparing the walls with a fingernail is a five-minute education that no paragraph replaces: the climb wall feels polished, the conventional wall shows the faint vertical striations of teeth that rubbed before they cut.
Reading and writing the modes deliberately
Two practical rituals make the knowledge operational. Reading: given any side-milling section of a program, determine the mode before running it. Find the comp call or, without comp, walk the coordinates: is the path going clockwise or counterclockwise, and which side is material? Pocketing cycles, profile loops, and arc moves all submit to the same five-second reconstruction. Doing this on inherited programs regularly turns up surprises, half-conventional pockets from an old CAM template, finish passes cut in the wrong direction, that explain longstanding grumbles about a job’s finish.
Writing: choose the mode per operation, on purpose. Roughing and finishing on a rigid machine: climb, path directions set accordingly. First pass over mill scale on hot-rolled plate: consider a conventional skin cut, then climb below the crust. Slotting is its own case, because a full-width slot is climb on one wall and conventional on the other simultaneously; where a slot’s finish matters, rough the slot undersize and finish each wall with its own climb pass. And on any machine whose backlash you do not know, a light test cut tells you more than the spec sheet: a machine that hums through a climb pass is fine; one that chatters or snatches has cast its vote.
The chip-load connection closes the loop: because conventional entry rubs before it cuts, running conventional at climb-calibrated feeds shortens tool life even when everything else is right, and the fix on such passes is typically a modest chip-load increase to get the tooth under the work-hardened layer its predecessor made. Feed arithmetic for either mode is the same feed-per-tooth calculation; what changes is the tolerance for thin chips.
Bottom line: the mode is in the direction
Climb milling enters thick and exits thin: better finish, longer tool life, the CNC default. Conventional enters thin and rubs: the choice for backlash-prone machines, abrasive skins, and deflection-sensitive moments. In code, direction is destiny: outside profiles clockwise or pockets counterclockwise are climb with forward spindle and right-hand tools, G41 is climb and G42 conventional under the same assumptions, and everything flips together if spindle hand or tool hand flips. Read the mode out of any program in five seconds, choose it per operation on purpose, and half the mysterious finish problems in a shop lose their mystery.
Frequently asked questions
What is the difference between climb and conventional milling?
In climb milling each tooth enters the cut at full chip thickness and exits thin, biting cleanly; in conventional milling the tooth enters at zero thickness and rubs before it cuts, thickening through the sweep. Climb gives better finish, longer tool life, and throws chips clear, making it the CNC default; conventional survives for backlash-prone machines, abrasive skins on castings and hot-rolled stock, and cuts where climb’s pull-in would deflect the tool or part.
How do I know if my G-code is climb or conventional milling?
Face the direction of travel and ask which side the material is on: with M03 and right-hand tooling, material on the right of travel is climb. Equivalent rules: outside profiles cut clockwise (viewed from above) are climb, inside pockets cut counterclockwise are climb, and with cutter compensation G41 is climb while G42 is conventional. All of it inverts together only if spindle direction or tool hand inverts.
Why is climb milling bad on machines with backlash?
Because climb pulls the cutter into the work along the feed direction. On a machine with play in its screws, that pull can snatch the table forward by the lash amount, and the cutter suddenly takes an oversized bite: gouged surfaces, broken tools, the classic manual-mill accident. Rigid CNC machines with preloaded ballscrews resist the pull, which is exactly why climb became the CNC default while conventional remained the manual-machine rule.
What is the best way to learn the codes and rules behind milling direction?
Drill the small rule set until it is reflex, G41 climb, G42 conventional, outside-CW and inside-CCW climb, all under M03 with right-hand tools, then practice the five-second reconstruction on real programs: travel direction, material side, mode. A free app like G-Code Sprint quizzes the comp and direction codes in short timed rounds and repeats what you miss, which keeps the mapping instant at the machine.
Should I ever mix climb and conventional in one part?
Routinely. A first pass over mill scale or casting skin often runs conventional to protect the edge, with climb below the crust. Slots are both modes at once full-width, so finish-critical slots get roughed undersize and each wall finished with its own climb pass. Deflection-sensitive finishing sometimes uses a conventional spring pass because its forces push away from the wall. The mode is an operation-level decision, not a part-level loyalty.