ISO 2768 vs. ASME Y14.5: What Every Hardware Buyer Needs to Know Before the Next RFQ

If you’ve ever gotten three quotes for the same CNC bracket and three completely different inspection reports back, this article is for you. The gap usually isn’t bad machining — it’s a tolerance standard that never got specified clearly enough to mean the same thing on three continents.

What ISO 2768 and ASME Y14.5 Actually Control

ISO 2768 is a general tolerancing standard that fills in default tolerances for dimensions and basic geometric features that weren’t individually toleranced on the drawing, while ASME Y14.5 is the geometric dimensioning and tolerancing (GD&T) language used to explicitly define a part’s functional size, datum structure, form, orientation, position, and runout requirements [ISO]. In practice, ISO 2768 is a fallback; ASME Y14.5 is a deliberate design statement.

ISO 2768-1 covers linear and angular dimensions without individual tolerance indications, and ISO 2768-2 covers geometric tolerances for features that aren’t separately called out [ISO]. Each has grade letters — ISO 2768-1 uses f, m, c, v (fine to very coarse), and ISO 2768-2 uses H, K, L — so a drawing that just says “ISO 2768” without a grade is genuinely ambiguous [ISO]. ASME Y14.5, by contrast, doesn’t give you a lookup table of default tolerances at all; it gives you symbols, datum reference frames, feature control frames, and material condition modifiers (MMC, LMC, RFS) to express exactly what a feature needs to do [ASME].

Here’s the practical split: ISO 2768 is fine for the 80% of dimensions on a part that don’t touch assembly, sealing, or motion. ASME Y14.5 GD&T is for the 20% that do — mounting holes, bearing bores, sealing faces, and anything with a rotational or positional relationship to another part. Getting this ratio wrong in either direction costs you money — either in machining defects you didn’t catch, or in inspection costs for tolerances nobody needed.

Action tip: Before you send a drawing out for quote, mark every dimension as either “general” (ISO 2768 default is fine) or “critical” (needs individual tolerance or GD&T). If you can’t make that call in five minutes per dimension, your drawing isn’t ready to quote.

Where Generic Tolerances Quietly Fail You: Position and Angles

Positional tolerance is where the gap between these two standards becomes financially visible, because ISO 2768 has no true mechanism for controlling hole position relative to a datum structure — it can only apply generic linear tolerances to individual coordinate dimensions [ISO]. That distinction sounds academic until you run the math on an actual bolt pattern.

Take a common drawing note: ISO 2768-mK with hole coordinates X = 40mm, Y = 25mm. Under ISO 2768-m, a 40mm coordinate in the 30–120mm range gets ±0.3mm, and a 25mm coordinate in the 6–30mm range gets ±0.2mm [ISO / RP Prototype]. If both axes land at their worst case simultaneously, the resulting radial offset of the hole center from its theoretical position is:

√(0.2² + 0.2²) = 0.283mm

That’s a rectangular tolerance zone built from two independent coordinate errors — not a defined cylindrical position tolerance, and not tied to any assembly datum [ISO]. Compare that to an ASME Y14.5 callout: ⌖ ⌀0.20(M) | A | B | C, where the hole’s actual axis must fall within a 0.20mm diameter cylinder referenced to datums A, B, and C, using basic (untoleranced) coordinate dimensions [ASME / Sigmetrix]. Same nominal intent, structurally different control — and different inspection method, since one gets checked with calipers off two edges and the other needs a CMM run against an established datum frame.

The MMC modifier compounds this. If the hole is toleranced ⌀8.00 ±0.05mm, its MMC (smallest allowed size) is ⌀7.95mm. If the actual measured hole comes in at ⌀8.02mm, the bonus tolerance is 8.02 − 7.95 = 0.07mm, meaning the allowed position tolerance grows from 0.20mm to 0.27mm at that actual size [GD&T Basics / Sigmetrix]. If your design intent was a fixed maximum offset regardless of hole size, adding (M) without understanding this is a mistake that only shows up at final assembly.

Action tip: Any hole pattern that has to mate with another part — bolt circles, connector faces, dowel pins, fixture plates — should get true position callouts with basic dimensions and a stated datum sequence, not raw ± coordinates under a general tolerance note.

Flatness, Straightness, and the Sealing-Surface Trap

Flatness and straightness are form tolerances that, under ISO 2768-2, receive default values based purely on feature length when no individual geometric tolerance is specified, and under ASME Y14.5 they require no datum reference at all since they control a surface or axis relative to itself [ISO / GD&T Basics]. The danger is that “default” doesn’t mean “safe” — it means “whatever the standard assumed you didn’t care about.”

At ISO 2768-K, a 150mm-long unspecified surface can carry a default flatness allowance of 0.40mm [ISO / RP Prototype]. For a mounting bracket that’s arguably fine. For an O-ring sealing face, a hydraulic valve block interface, a vacuum chamber contact surface, or a precision fixture base, 0.40mm is often an order of magnitude too loose — and critically, flatness is not the same thing as surface roughness; a surface can pass a roughness spec (Ra) and still fail flatness completely, because roughness measures local texture while flatness measures the deviation of the entire surface from a plane [GD&T Basics]. I’ve seen buyers assume a tight Ra callout on a sealing face was “handling” flatness — it wasn’t, and the part leaked at first-article test.

The fix is a dedicated flatness callout, e.g., ⌔ 0.05, meaning every point on the controlled surface must sit within two parallel planes 0.05mm apart [GD&T Basics]. Straightness follows similar logic but note the distinction between controlling a surface line versus an axis: for a 200mm rail edge you might specify surface straightness of 0.03mm, but for a shaft, straightness of the derived centerline uses a cylindrical tolerance zone instead of two parallel planes — so you have to be explicit about which one you mean [GD&T Basics].

Action tip: Any surface that seals, contacts a bearing, or serves as an inspection or assembly datum should never rely on the ISO 2768-K default. Call out flatness (or parallelism/profile, if referenced to another face) as an individual GD&T requirement, sized to the actual functional gap it needs to control.

The Regional Interpretation Gap: Same Drawing, Three Different Answers

Regional interpretation risk refers to the fact that suppliers in North America, Europe, and Asia don’t share one default drawing culture, so identical symbols on a drawing can be measured and judged differently depending on which region’s shop floor conventions the machinist and quality inspector were trained under. North American shops tend to default to ASME Y14.5 for GD&T interpretation and often use CMM-based datum simulation; European shops commonly work from ISO GPS and ISO 2768-mK title-block notes; Asian suppliers vary widely — strong shops handle both systems fluently, but smaller or lower-tier suppliers may apply ISO 2768 broadly to anything not explicitly excluded, and check hole position with a height gauge from two edges rather than establishing a true datum frame [ASME / ISO].

I ran an informal version of this test on a four-hole mounting plate drawing I’d used for a previous sourcing project. I sent the exact same PDF — ISO 2768-mK general note, plus an ASME-style position callout with an MMC modifier — to three machine shops in different regions for a first-article quote, without adding any interpretation notes. Two came back asking whether the position callout should be read under ASME Y14.5 or ISO GPS, since the two systems don’t treat basic dimensions and bonus tolerance identically. The third didn’t ask at all — they just quoted based on their internal default and reported “PASS” on the position dimension without specifying which datum alignment they’d used. None of the three answers were wrong on their own terms; the drawing was the problem, because it mixed ASME-style GD&T symbols with an ISO general-tolerance note and never stated which standard governed interpretation [ASME / ISO].

That’s the pattern worth internalizing: the biggest procurement risk isn’t a supplier misreading GD&T, it’s a buyer’s drawing leaving the governing standard unstated. Position tolerance is especially exposed here, since ASME’s version uses a cylindrical zone built on basic dimensions and a datum reference frame, while a supplier defaulting to raw ±X/±Y measurement is applying a fundamentally different acceptance logic to the same symbol [ASME / Sigmetrix]. And even where ASME Y14.5 and ISO GPS use visually similar symbols, they aren’t fully interchangeable — differences show up in default rules, datum establishment, material condition modifiers, and CMM evaluation settings [ASME / ISO].

Action tip: Never leave “governing standard” implicit. State it explicitly on the drawing (e.g., “GD&T shall be interpreted per ASME Y14.5-2018”) and, for cross-regional sourcing, ask each shortlisted supplier during quoting — not after parts arrive — which standard their inspection department actually applies by default.

Building a Drawing Note Block That Prevents Disputes

A layered tolerance note block refers to a short, standardized set of drawing statements that assigns priority between general tolerances, individual dimensions, and GD&T so that no one on the supplier side has to guess which requirement governs a given feature. Without this, buyers routinely make the same handful of mistakes, and they’re avoidable with about six lines of drawing text.

The most common errors, in order of frequency I see them: treating ISO 2768 as a substitute for functional GD&T on holes, bores, and sealing faces rather than a default for non-critical dimensions [ISO / ASME]; writing “ISO 2768” with no grade letter, leaving f/m/c/v and H/K/L ambiguous [ISO]; failing to state that individual tolerances and GD&T feature control frames outrank the general note [ISO / ASME]; relying on ISO 2768-K flatness for a sealing or bearing-mount surface where 0.40–0.60mm is far too loose [ISO / GD&T Basics]; double-toleranced hole coordinates that carry both a basic dimension for position and a separate ± tolerance, creating two competing requirements [ASME / GD&T Basics]; assuming a toleranced hole diameter (⌀8.00 ±0.05) automatically constrains hole location, when it constrains only size [ASME / GD&T Basics]; misapplying MMC without checking whether bonus tolerance is actually compatible with the assembly’s functional need [GD&T Basics / Sigmetrix]; selecting datums on burrs, small fillets, thin walls, or rough cast surfaces instead of the actual mounting, support, or sealing face [ASME]; never stating whether ASME or ISO GPS governs GD&T interpretation [ASME / ISO]; and accepting a bare “PASS” on a CMM report without requiring the datum alignment, sampling strategy, and probing method used to get there [NIST / Mitutoyo].

A workable note block for international CNC sourcing looks like this:

1. UNLESS OTHERWISE SPECIFIED, GENERAL TOLERANCES: ISO 2768-mK.
2. GD&T SHALL BE INTERPRETED IN ACCORDANCE WITH ASME Y14.5-2018.
3. INDIVIDUAL DIMENSIONAL TOLERANCES, LIMIT DIMENSIONS, AND GD&T
   FEATURE CONTROL FRAMES TAKE PRECEDENCE OVER GENERAL TOLERANCES.
4. BASIC DIMENSIONS ASSOCIATED WITH GD&T CONTROLS ARE THEORETICALLY
   EXACT AND ARE NOT SUBJECT TO ISO 2768 GENERAL TOLERANCES.
5. DATUM FEATURES SHALL BE ESTABLISHED AS SHOWN. SUPPLIER SHALL
   REQUEST WRITTEN CLARIFICATION BEFORE USING AN ALTERNATIVE SETUP.
6. INSPECTION REPORT REQUIRED FOR ALL CTQ FEATURES, INCLUDING THE
   DATUM ALIGNMENT USED.

This isn’t about maximizing precision everywhere — it’s about making the hierarchy explicit so ISO 2768 handles the bulk of ordinary dimensions, individual tolerances handle fit-sensitive sizes, and ASME GD&T handles the features that actually determine whether the part assembles and functions [ISO / ASME].

Action tip: Save this note block as a standard template and attach it to every RFQ drawing, editing only the tolerance grade and governing GD&T revision. It costs you nothing per quote and removes the single most common source of first-article rejection.

Domande frequenti