Set Screw Dimensions Calculator

Point dimensions for all four ISO socket set screws — cup (4029), cone (4027), flat (4026) and dog (4028) — with an axial holding power and the torque it will hold on your shaft, the point-type and multiple-screw factors that change it, and the selection logic that actually decides which point you want. The holding power is a constant-pressure fit to a named maker’s published inch table and the fit’s residuals are printed rather than hidden.

Last updated 24 Sep 2026

Set screw dimensions

Size and point → dimensions and holding power

Set screw threadISO 4026 to 4029 cover M1.6 to M24; the sizes here are the ones a shaft collar or a pulley actually uses.

Point typeThe selection logic is the page: cup for a permanent hold on a shaft it is allowed to mark, cone for the highest holding into a spotted seat, flat for repeated adjustment on a hardened shaft, dog into a matching flat or hole.

How many screws, and whereMachine Design’s figures: a second screw in line along the axis roughly doubles the holding power, 60° apart gives 1.75 times, and diametrically opposed gives only 1.3 — because opposed screws push against each other instead of adding grip.

Shaft diameter (mm)Used to turn the axial holding power into a transmissible torque: Safety Socket’s own relation is torsional holding power = axial holding power × shaft radius.

Torque you need to transmit (N·m)What the joint has to carry. The headline tells you what the screw will hold; this tells you whether that is enough.

Not a circuit: the four ISO point types at the same thread size, to a scale normalised by the thread diameter, with the mark each one leaves in the shaft drawn underneath the shaft surface. All four bodies are the same because the thread is; the four bottom ends are the whole difference between the standards. Left to right: the CUP's annular edge, which bites a ring into the shaft and is the mechanism as well as the damage; the CONE truncated to dt, which is about half the flat's diameter and drives a single deep crater; the FLAT, a full circle that barely marks anything and holds by friction alone; and the DOG, a cylindrical extension that has to enter a matching flat or hole to depth z — drawn here as the pocket it needs, because landing it on a plain round shaft is the one clear mistake on this page. The box picks out the type you have selected. The bar at the bottom is the combined point and screw-count factor from zero to four times, with a tick at 1×, so the second screw's contribution is visible against the point's: changing the point moves it by at most 7%, and putting a second screw 60° round moves it by 75%.

6,248NExample

An M8 cup-point set screw on a 20 mm shaft, one screw, with 25 N·m to transmit

A fitted constant pressure, two published factors, and a radius

AHP ≈ k·d² with k = 97.62 N/mm² (14,159 lbf/in²) · AHP = AHP_cup × point factor × screw-count factor · torque held = AHP × shaft radius

k: a constant pressure on the nominal thread area, fitted by least squares to Safety Socket’s published axial holding powers over the sizes 1/4 in and up. It is a DESCRIPTION of one maker’s table, not a standard, and the published values sit between 6% below and 16% above it
d: nominal screw diameter. The d² is not an assumption forced on the data: fitted freely over the whole table the best exponent is 2.10
point factor: Unbrako’s multipliers on the cup-point figure: cone 1.07, flat and dog 0.92, oval 0.90. The cone’s advantage is small, which is worth knowing before cutting a spotted seat for it
screw-count factor: Machine Design’s: 2.0 for two in line along the axis, 1.75 at 60° apart, 1.30 diametrically opposed. The geometry of where the second screw goes is worth more than the second screw
torque held: Safety Socket’s own relation — torsional holding power is the axial figure times the shaft radius. It is the tangential force at the shaft surface doing the work

Worked example

An M8 cup-point set screw on a 20 mm shaft, one screw, with 25 N·m to transmit
Estimated axial holding power = k·d² = 97.62 × 8² = 6,248 N, which is 1,405 lbf. For scale, the published inch table's 5/16 in (7.94 mm) row is 1,600 lb = 7,117 N, so the estimate is in the right place
Cup point, so the point factor is 1.00, and one screw, so the count factor is 1.00 — the headline is the raw figure
Torque it will hold = 6,248 N × 0.010 m = 62.5 N·m on a 20 mm shaft. Against the 25 N·m asked for that is 2.50 times — sound, with no great margin, and this is a typical strength with no safety factor in it
The cup's own contact is an annulus of diameter dz = 5.0 mm, which is 25% of the shaft diameter — that is the size of the permanent mark it will leave
Switch to a cone point and the factor is 1.07, so 6,685 N: a 7% gain for the cost of spotting the shaft. Add a second screw 60° round instead and the factor is 1.75, giving 10,934 N — the second screw is worth ten times what the point change is. Put that second screw diametrically opposite instead and you get only 1.30, or 8,122 N
And the thing the arithmetic cannot say: on a PLAIN round shaft this is a slipping joint whatever the number says. Under reversing torque it will creep. A flat, a groove or a key changes the question

Which point, and what it does to the shaft

This table is the page. The arithmetic above it is a one-line multiplication; the choice of point is the decision. Two things are worth reading twice. The cone point’s advantage is only 7% — Unbrako’s own multiplier — so “deep penetration gives the highest holding power” is true and small, and the reason to choose a cone is usually positive location in a spotted seat rather than raw grip. And the dog point is the only one of the four that is not primarily a friction joint: with a matching flat it is a key, and Machine Design note that point penetration “contributes as much as 15% to the total holding power”, so the rest is friction from the clamp force either way. This page sizes a part; it does not certify one. Where the answer carries a consequence — a load path, a lifting duty, a pressure boundary, a fastener holding something that can fall — confirm it against the design code that governs the application, and against the manufacturer’s own rating, before relying on it.
Point	Use it when	What it does to the shaft	Holding power
Cup — ISO 4029 / DIN 916	The default, and the right default. Permanent or semi-permanent location where the shaft may be marked. Unbrako: “use against hardened shafts, in zinc, die castings and other soft materials where high tightening torques are impractical”	MARKS IT. The annular edge bites in and raises a burr, and the mark is permanent. That is how it holds — it makes its own detent — and it is also why a shaft that has to slide through a bearing afterwards should not be held with one. A knurled cup adds a serrated locking action against vibration	Baseline, 1.00
Cone — ISO 4027 / DIN 914	The highest holding available, into a SPOTTED SEAT drilled or milled to match. Unbrako: “for permanent location of parts. Deep penetration gives highest axial and holding power”	Deep single-point penetration. Without a spotted seat it drives its own crater, and on a hard shaft it will not. The angle matters: ISO 4027 is 120° on short screws and 90° on longer ones, so a seat cut for one is wrong for the other. The usual instruction is to spot the shaft to half the point’s depth	1.07 — only 7% above cup, which surprises people
Flat — ISO 4026 / DIN 913	Repeated adjustment. Unbrako: “use where parts must be frequently re-set, as it causes little or no damage”. Also the choice on a hardened or ground shaft you must not mark	Very little. A full circular face at dp, no penetration to speak of, so the holding is friction alone	0.92
Dog — ISO 4028 / DIN 915	Where the screw has to LOCATE as well as hold: it enters a matching flat, groove or hole and becomes a key. Also for permanent setting of a part that must not creep	Nothing, if the flat or hole is there. Everything, if it is not — a dog point landing on a round shaft bears on its edge and will gall. The flat or hole must be at least z deep, and z is tabulated above in two forms	0.92, but the located version is not really a friction joint at all
Oval — not in the ISO 4026–4029 set	Frequent adjustment against a curved or angled surface. Unbrako: “use for frequent adjustment without deformation of part it bears against”	Least of all, which is the point	0.90

Point dimensions, all four ISO standards at once

One thread, four contacts. The cup is an annular edge of diameter dz that bites in; the flat is a full circle of diameter dp that does not; the cone is truncated to dt, which is roughly HALF the flat’s diameter, so all the load goes through a small area and penetration is deep; and the dog is a cylindrical extension of diameter dp and length z that has to enter something. The last column is the reason a cone point holds best and damages most: the same force through a quarter of the area. The two dog lengths matter on a drawing — the flat or hole in the shaft has to be at least z deep, and the long form is roughly double the short. ISO 4027’s own note is that the cone angle is 120° on short screws and 90° on longer ones, which is why a spotted seat has to be cut to match the screw you are actually using. These dimensions come from a published standard’s table, not from a formula. The standard itself is cited below and the printed values are attributed to the catalogue they were taken from; a different publisher may round differently in the last digit.
Thread	Cup dz max (ISO 4029)	Flat/dog dp max (4026/4028)	Cone truncation dt max (4027)	Dog z short (mm)	Dog z long (mm)	Cup dz ÷ thread	Cone dt ÷ flat dp
M3	1.40	2.00	0.75	1.00	1.75	0.467	0.375
M4	2.00	2.50	1.00	1.25	2.25	0.500	0.400
M5	2.50	3.50	1.25	1.50	2.75	0.500	0.357
M6	3.00	4.00	1.50	1.75	3.25	0.500	0.375
M8	5.00	5.50	2.00	2.25	4.30	0.625	0.364
M10	6.00	7.00	2.50	2.75	5.30	0.600	0.357
M12	8.00	8.50	3.00	3.25	6.30	0.667	0.353
M16	10.00	12.00	4.00	4.30	8.36	0.625	0.333
M20	14.00	15.00	5.00	5.30	10.36	0.700	0.333
M24	16.00	18.00	6.00	6.30	12.43	0.667	0.333

Two screws, and where the second one goes

The published figures are Machine Design’s: holding power “is approximately doubled when the second screw is installed in an axial line with the first”, is “only about 30% greater when the screws are diametrically opposed”, and a 60° displacement “gives 1.75 times the holding power of one screw”. Note that the commonly repeated figure of 120° is NOT what this source says: the recommended circumferential displacement is 60°, and 120° would sit between the 60° and 180° cases. Mutual Screw’s own reference states the dependence without numbers: “when two set screws are used in a set screw collar, their holding power is determined by their location with respect to each other.” Unbrako publish the same relationship as a chart of holding power against the angle between screws. This page sizes a part; it does not certify one. Where the answer carries a consequence — a load path, a lifting duty, a pressure boundary, a fastener holding something that can fall — confirm it against the design code that governs the application, and against the manufacturer’s own rating, before relying on it.
Arrangement	Holding power against one screw	Why
Two in line along the shaft axis	About 2.0×	Two independent contacts on the same generator of the shaft, each carrying its own normal force. The nearest thing to simply adding two screws together.
Two displaced 60° around the shaft	1.75×	Machine Design call this “the best compromise between maximum holding power and minimum metal between tapped holes” — you get most of the doubling and the two tapped holes still have material between them. This is the practical recommendation.
Two diametrically opposed, 180° apart	Only about 1.3×	The two normal forces are nearly collinear and opposite, so the second screw mostly squeezes the shaft rather than adding grip. It is also the arrangement people reach for first, because it looks balanced.
One screw on a plain round shaft	1.0× — and it is a slipping joint	A single point contact on a cylinder, holding by friction and by whatever detent it has dug for itself. It will creep under reversing torque. Where the load matters, add a second screw or put a flat, a groove or a key on the shaft.

Safety Socket’s published seating torque and axial holding power

This is a named manufacturer’s table and it is the source for every holding-power figure on the page. Two things follow from it. The last column is a least-squares fit of AHP = k·d² over the sizes 1/4 in and up, and k comes out at 14,159 lbf/in² — that is, the table is very nearly a constant PRESSURE on the nominal thread area, which is what you would expect if the cup’s edge is bearing into the shaft. Over those nine sizes the published values sit between 6% below and 16% above the fit. Fitted freely over the whole table from No. 0 the best exponent is 2.10, not 2, and the spread widens to −26% to +16%: the small sizes do not follow the law as well. The metric estimate on this page uses the 1/4-in-and-up constant, and the chart above shows you the residuals rather than hiding them. Note also the stainless column: a stainless set screw takes 10% to 20% less seating torque, so it holds proportionally less. These dimensions come from a published standard’s table, not from a formula. The standard itself is cited below and the printed values are attributed to the catalogue they were taken from; a different publisher may round differently in the last digit.
Size	d (in)	d (mm)	Seating torque, alloy (in·lb)	Seating torque, stainless (in·lb)	Axial holding power (lb)	… in newtons	Published ÷ the fitted d² law
No. 0	0.0600	1.52	0.86	0.40	50	222	0.981
No. 1	0.0730	1.85	1.80	1.20	63	280	0.835
No. 2	0.0860	2.18	1.80	1.20	80	356	0.764
No. 3	0.0990	2.51	5.00	4.00	110	489	0.793
No. 4	0.1120	2.84	5.00	4.00	140	623	0.788
No. 5	0.1250	3.17	9.50	7.00	180	801	0.814
No. 6	0.1380	3.51	9.50	7.00	200	890	0.742
No. 8	0.1640	4.17	19.40	16.00	365	1,624	0.958
No. 10	0.1900	4.83	33.50	26.00	550	2,447	1.076
1/4 in	0.2500	6.35	77.90	70.00	1,000	4,448	1.130
5/16 in	0.3125	7.94	156.00	130.00	1,600	7,117	1.157
3/8 in	0.3750	9.52	273.00	230.00	2,200	9,786	1.105
7/16 in	0.4375	11.11	428.00	340.00	3,000	13,345	1.107
1/2 in	0.5000	12.70	615.00	500.00	3,700	16,458	1.045
5/8 in	0.6250	15.88	1,315.00	980.00	5,600	24,910	1.013
3/4 in	0.7500	19.05	2,150.00	1,700.00	7,500	33,362	0.942
7/8 in	0.8750	22.22	5,130.00	4,000.00	10,500	46,706	0.969
1 in	1.0000	25.40	7,010.00	5,600.00	14,500	64,499	1.024

What this page leaves out

The first exclusion is deliberate and structural: this site already has a hex key and socket size page, and a set screw page that reprinted the socket column would be competing with it. The second and third are about honesty: one published table was unreadable, and the metric figures do not exist in public, so the page fits, states the fit, and prints the error.
What	Why
The hex socket size and the key that fits it	There is a dedicated page for hex key and socket sizes already, and duplicating it would split the traffic between two pages answering the same question. The socket size is in every one of the four ISO standards’ tables; this page prints the POINT dimensions, which is the part that touches the shaft.
Unbrako’s own holding-power table	It could not be extracted without row shifts. Three attempts returned values that did not reconcile with each other — one of them by a factor of five — and a holding-power figure read out of a shifted row is worse than no figure. Safety Socket’s table is used instead, whole and attributed, and the point-style multipliers and the selection wording are taken from Unbrako’s guide, which extracted cleanly.
A metric published holding-power table	None was found. The estimate here is a constant-pressure fit to the inch table, and the fit’s residuals are printed in the chart and the table rather than hidden. If you need a guaranteed figure for a metric screw, it has to come from the maker.
Thread stripping in the collar	A set screw’s limit is almost never its own thread — it is the shaft surface, the seating torque a small hex socket can take, or the collar’s thread in a soft material. Thread engagement is handled properly in the bolted-joint cluster.
Seating torque for a specific lubricant or coating	Seating torque is a friction question and the published figures are for the plain product. Unbrako’s own caution applies to all of this: “tabulated axial and torsional holding powers are typical strengths and should be used accordingly, with specific safety factors appropriate to the given application and load conditions.”

The point is the decision; the second screw is the lever

Anyone searching for set screw dimensions wants one of two things: the point diameter, so the drawing can show what it does to the shaft, or the holding power, so they can decide whether one screw is enough. The point dimensions are exact and come straight from the four ISO standards’ tables. The holding power does not exist for metric screws in any public source this batch could find, so this page does something explicit about it instead of quietly making a number up.

Here is what it does. Safety Socket publish a complete inch table — nominal size, seating torque for alloy steel and for stainless, and axial holding power in pounds, from No. 0 to 1 in. Fit AHP = k·d² to it by least squares over the sizes 1/4 in and up and k comes out at 14,159 lbf/in²: the table is very nearly a constant PRESSURE on the nominal thread area, which is what you would expect if what is happening is a cup edge bearing into the shaft. Over those nine sizes the published values sit between 6% below and 16% above the fit. Fitted freely over the whole table from No. 0 the best exponent is 2.10 rather than 2, and the spread widens to −26% to +16% — the small sizes do not follow the law as well. The chart on this page plots the published points and the fit together so the residual is visible. That is the accuracy of every holding-power figure here, and Unbrako’s own caution applies on top of it: these are “typical strengths … with specific safety factors appropriate to the given application and load conditions”.

The selection logic is the part of this page that is worth keeping, and it is not arithmetic. Cup point is the default and the right default: it marks the shaft permanently, and the mark is the mechanism — the annular edge digs a detent for itself. Unbrako’s own wording is that it is for use “against hardened shafts, in zinc, die castings and other soft materials where high tightening torques are impractical”. Cone point gives the deepest penetration and the highest holding, into a seat spotted to match — and the surprise is how small the gain is: Unbrako’s multiplier is 1.07, so 7%. Flat point is for repeated adjustment and for a hardened or ground shaft you must not mark, at a cost of 8%. And dog point is different in kind: it enters a milled flat, a groove or a hole and becomes a key, which is a positively located joint rather than a friction one — but a dog point landed on a plain round shaft bears on the edge of its own cylindrical extension and galls, so it is the one arrangement on this page that is simply wrong.

The cone point’s angle is not one number, which catches people. ISO 4027 specifies 120° on short screws and 90° on longer ones — the boundary is a step line in the standard’s own length table — so a shaft spotted for a 90° cone will not seat a 120° one. The screw then bears on a circle rather than a cone, which is the same failure as a countersunk screw in the wrong-angle recess, and the geometry of that is worked out on the counterbore and countersink calculator.

And the biggest lever is the second screw, not the point. Machine Design publish the geometry: a second screw in line along the shaft axis roughly doubles the holding power; displaced 60° around the shaft it gives 1.75 times a single screw, which they describe as “the best compromise between maximum holding power and minimum metal between tapped holes”; and diametrically opposed — the arrangement that looks most balanced and that people reach for first — gives only about 30% more, because two opposed normal forces squeeze the shaft instead of adding grip. Note that the widely repeated figure of 120° is not what this source says; the recommended circumferential displacement is 60°, and 120° sits between the 60° and 180° cases. Either way, a single screw on a plain round shaft is a slipping joint: it holds by friction and by the crater it has dug, and under reversing torque it creeps. Where the torque matters, put a flat on the shaft, add a second screw, or use a key. What this page does not print is the hex socket size — this site already has a page for hex key and socket sizes and there is no sense in having two. The shaft and collar end of the same problem is on the dowel pin and hole calculator if what you actually need is location rather than grip, and the retaining ring groove calculator if what you need is axial restraint rather than torque.

Frequently asked questions

Which set screw point should I use?

Cup point unless something rules it out — it is the default, it holds well, and the permanent mark it leaves on the shaft is actually the mechanism. Use a FLAT point where the shaft must not be marked or the screw will be re-set often: Unbrako say “use where parts must be frequently re-set, as it causes little or no damage”. Use a CONE point into a spotted seat where you want the deepest penetration and positive location — but know that the holding-power gain is only 7%. Use a DOG point where the shaft has a matching flat, groove or hole, because then the screw is a key rather than a friction grip. Never land a dog point on a plain round shaft.

How much torque will one set screw hold?

Axial holding power times the shaft radius — that is Safety Socket’s own relation. This page estimates an M8 cup point at about 6,250 N of axial holding power, which on a 20 mm shaft is about 62 N·m. Treat that as an order of magnitude with a stated error, not a rating: it comes from a constant-pressure fit to one maker’s inch table whose published values sit within −6% to +16% of the fit, and the maker calls them typical strengths with no safety factor included. It also assumes a clean dry shaft and full seating torque.

Is one set screw enough?

On a plain round shaft under steady torque, sometimes. Under reversing or shock torque, no — it is a friction joint on a cylinder and it will creep. The published remedies, in order of what they buy: a second screw in line along the axis (about double), a second screw 60° round the shaft (1.75 times, and described as the best compromise because the tapped holes still have metal between them), a larger screw (holding goes as the diameter squared), or a milled flat for a dog point. Two screws diametrically opposed are worth only about 1.3 times a single one, which is the opposite of what most people assume.

Should two set screws be 120° apart?

The source this page could verify says 60°, not 120°. Machine Design give three figures: two screws in an axial line about double a single screw, 60° apart 1.75 times, and diametrically opposed only about 1.3 times — and they name 60° as “the best compromise between maximum holding power and minimum metal between tapped holes”. On that curve 120° sits between the 60° and 180° cases, so it is better than opposed and worse than 60°. Unbrako publish the same relationship as a chart of holding power against the angle between screws. If you have seen 120° quoted as the recommended practice, it is worth checking against the source, because this batch could not find one that says it.

Will a set screw damage the shaft?

A cup point will, permanently, and that is by design: the annular edge bites in and raises a burr, making its own detent. The mark’s diameter is dz, which this page prints as a percentage of the shaft. A cone point damages more, in one deep spot. A flat point damages least of the three and holds least. If the shaft has to pass through a bearing, a bush or a seal after the collar comes off, the burr is the problem rather than the dent, and the answer is a flat point or a milled flat with a dog point.

Why does this page not give the hex socket size?

Because this site already has a page for hex key and hex socket sizes, and two pages answering the same question compete with each other. The socket size is in every one of the four ISO standards’ tables. What this page prints instead is the POINT dimension, which is the end that touches the shaft and the thing a drawing has to show.

Do stainless set screws hold as well as alloy steel ones?

No, and the published seating torques show why. Safety Socket list stainless seating torques 10% to 20% below the alloy-steel figures at every size — 70 in·lb against 77.9 at 1/4 in, 500 against 615 at 1/2 in. Since Unbrako state that “holding power is almost directly proportional to seating torque” for a given size and point, the holding power follows the torque down. Stainless also galls more readily in its own thread, which is a separate problem with the same cure: a lubricant or an anti-seize.

Related calculators

References

Fuller Fasteners. ISO 4026 / 4027 / 4028 / 4029 Specifications — Hex Socket Set Screws with Flat, Cone, Dog and Cup Point. Source for every point dimension used here: the flat and dog point diameter dp, the cup point diameter dz, the cone truncation diameter dt, and the short and long dog point lengths z. Its ISO 4027 page also carries the angle note — 120° above the step line in the length table and 90° below it — which is why a cone point on a short screw is a different cone from one on a long screw.
Safety Socket Screw Corporation. Set screw tightening torque. The one complete, internally consistent set-screw table this batch reached: nominal diameter, hex key size, alloy-steel and stainless seating torque in inch-pounds, and axial holding power in pounds, from No. 0 to 1 in. Also the relation “For Torsional Holding Power (THP), Multiply Axial Holding Power (AHP) by Shaft Radius (R)”.
Unbrako. Socket Products Engineering Guide. Source for the point-style selection logic, quoted: cup point “use against hardened shafts, in zinc, die castings and other soft materials where high tightening torques are impractical”; cone point “for permanent location of parts. Deep penetration gives highest axial and holding power”; flat point “use where parts must be frequently re-set, as it causes little or no damage”; oval point “use for frequent adjustment without deformation of part it bears against”. Also the point-style multipliers on the cup-point figure — cone 1.07, flat and dog 0.92, oval 0.90 — the statement that “holding power is almost directly proportional to seating torque in a cup, flat, and oval point screws”, and the caution that “tabulated axial and torsional holding powers are typical strengths and should be used accordingly, with specific safety factors appropriate to the given application and load conditions”. Its own holding-power TABLE could not be extracted without row shifts and is therefore NOT reproduced here.
Machine Design, Setscrews (machinedesign.com, article 21834719). The multiple-screw figures used here: holding power “is approximately doubled when the second screw is installed in an axial line with the first”, is “only about 30% greater when the screws are diametrically opposed”, and a 60° displacement “gives 1.75 times the holding power of one screw” — described as “the best compromise between maximum holding power and minimum metal between tapped holes”. Also that “setscrew point penetration contributes as much as 15% to the total holding power”. Mutual Screw & Supply’s socket set screw reference states the same dependence in words: “when two set screws are used in a set screw collar, their holding power is determined by their location with respect to each other.”