Bolt Grade Markings and Why Torque Specs Differ

Reading the markings

Metric fasteners carry their property class stamped on the head - 8.8, 10.9, 12.9 - beside the manufacturer's mark. Under ISO 898-1 the first number is the nominal tensile strength in hundreds of megapascals, and the two multiplied give one tenth of the nominal yield: class 8.8 reads as roughly 800 MPa tensile, 640 MPa yield.

Treat that as a mnemonic, not a specification. The actual minima in ISO 898-1:2013 Table 3 sit at or slightly above what the designation implies: class 10.9 is specified at 1040 MPa tensile and 940 MPa yield, not the 1000 and 900 the name suggests, and class 12.9 at 1220 and 1100. Where a strength value matters, read the table.

Inch fasteners under SAE J429 use radial lines instead. Grade 2 has no marks, Grade 5 has three, Grade 8 has six. The absence of marks is genuinely ambiguous - an unmarked bolt might be a Grade 2, or something with no defined strength at all, and inspection cannot tell them apart. Unmarked hardware does not belong in a loaded joint.

Nuts are classified separately under ISO 898-2, and the joint takes the strength of the weaker half: a class 10.9 bolt in a lower-class nut strips the nut first.

Strength is banded by diameter - this is the part charts get wrong

Neither standard gives one strength figure per grade. Both tabulate strength against nominal diameter, and the bands are not decorative.

Under ISO 898-1:2013, class 8.8 is the banded class in common use: 580 MPa proof, 640 yield and 800 tensile at 16 mm and below, rising to 600, 660 and 830 above 16 mm. Classes 10.9 and 12.9 hold one set of values - 830/940/1040 and 970/1100/1220 MPa - across the range. The tabulated scope starts at M1.6; below that no property class is defined and no proof load should be quoted.

Under SAE J429 the banding is far more aggressive, and it steps down rather than up. Grade 2 runs at 55 ksi proof from 1/4 through 3/4 inch and drops to 33 ksi above 3/4 inch - a 40% fall. Grade 5 runs at 85 ksi from 1/4 through 1 inch and drops to 74 ksi above 1 inch. Grade 8 alone is flat, at 120 ksi across the whole 1/4 to 1-1/2 inch range.

J429 also has a hard boundary at each end: no grade for numbered sizes below 1/4 inch, and none above 1-1/2 inch. That is why ThreadRef publishes no proof load or torque for a #10-24 or a 2-4.5 UNC. Fasteners that large are normally specified to ASTM A193, A320 or an equivalent. A chart printing a Grade 8 torque for a 2 inch bolt is inventing it.

• ISO 898-1 class 8.8: 580 MPa proof at d ≤ 16 mm, 600 MPa at d > 16 mm.
• ISO 898-1 classes 10.9 and 12.9: one value across the range.
• SAE J429 Grade 2: 55 ksi proof to 3/4 in, then 33 ksi.
• SAE J429 Grade 5: 85 ksi proof to 1 in, then 74 ksi.
• SAE J429 Grade 8: 120 ksi proof throughout its 1/4-1-1/2 in scope.

What the banding does to real numbers

The clearest demonstration is a Grade 2 bolt either side of the 3/4 inch boundary. A 3/4-10 UNC has a tensile stress area of 0.3345 in² in the 55 ksi band, giving a target preload of 13,797 lbf at 75% of proof. A 1-8 UNC has a stress area of 0.6057 in² - 81% larger - but sits in the 33 ksi band, so its preload is 14,992 lbf. Nearly double the metal, under 9% more clamp force. A chart carrying 55 ksi across both would overstate the 1 inch bolt by two thirds.

The Grade 5 boundary at 1 inch does the same one size later, and it changes how the grades relate. At 1/2-13 UNC, Grade 8 gives about 41% more preload than Grade 5 (12,771 lbf against 9,046). At 1-1/8-7 UNC, where Grade 5 has dropped into its 74 ksi band and Grade 8 has not moved, the gap widens to about 62% (68,695 against 42,362). The premium for going up a grade is not a fixed percentage.

The metric step runs the other way and is gentler. At M16, class 8.8 is still in its 580 MPa band: 68.15 kN of preload against 97.53 kN for class 10.9, a ratio of 1.43. At M20, class 8.8 has stepped up to 600 MPa and the ratio narrows to 1.38 - small, but real, and the reason every size page states which band its figures came from.

How a torque figure is actually built

Torque is not a strength property. It is an indirect, and rather poor, way of achieving preload - the tension locked into the bolt when the joint is tightened. Preload holds the joint together, resists fatigue and stops the fastener loosening. Torque is only the handle we have on it.

The chain runs in four steps. Take the proof strength for the grade at that diameter band. Multiply by the tensile stress area for the proof load. Take a fraction of that as the preload target - 75% is the usual convention. Then convert preload to torque with T = K · F · d, where d is the nominal diameter and K the nut factor.

Every step is a decision, not a constant. Change the preload fraction and every number moves with it. Cross a diameter band and the proof strength moves. Change the pitch and the stress area moves. Change the friction and K moves - which is where most chart disagreement lives.

The nut factor is where torque specs diverge

K is a lumped friction coefficient covering the threads and the bearing face under the head or nut. It is not derived from anything - it is an empirical figure for a particular combination of material, finish and lubricant. ThreadRef defaults to 0.20 for dry, as-received steel, with 0.22 zinc-plated, 0.18 lightly oiled and 0.12 for wax, moly or anti-seize.

Because torque scales linearly with K, those values are not a footnote. Torquing a moly-lubricated bolt to a figure calculated for dry K = 0.20 delivers roughly 1.67 times the intended preload - enough to take a fastener past yield. The error runs both ways: a lubricated-K figure on a dry, rusty bolt leaves the joint badly under-clamped. This one variable explains more chart disagreement than everything else combined.

Even with friction fixed, real preload from torque control commonly scatters by ±25-30%. That is not a flaw in the equation; it is what friction does. It is why critical joints use angle control, torque-to-yield, or direct measurement of bolt elongation rather than torque alone.

• K = 0.20 dry as-received steel - the usual default.
• K = 0.22 zinc-plated and unlubricated.
• K = 0.18 lightly oiled.
• K = 0.12 wax, moly or anti-seize - verify against the product data.
• Torque is directly proportional to K: halving friction nearly doubles preload at the same torque.

The other reasons two specs disagree

Friction is the largest variable but not the only one. The preload target is itself a choice: 75% of proof load is a common convention, some specifications work to a percentage of yield, and torque-to-yield procedures deliberately take the fastener past it and require replacement after one use.

The joint's own materials frequently override the fastener's capability. A steel bolt into a tapped aluminium casting is limited by the casting, and the correct torque may be well below anything a fastener chart suggests. Gaskets creep, plastics embed, thread-locking compounds and lock nuts add running friction that is not what K represents, and reused fasteners no longer behave like new ones.

Then there are the sizes where no general figure exists. Pipe threads seal against the mating flanks rather than being tightened to a preload, so they are specified by turns past hand-tight. Numbered inch sizes and anything above 1-1/2 inch fall outside J429. Below M1.6 no ISO 898-1 property class is defined.

So: use a general torque estimate to sanity-check a figure or size a wrench, and the manufacturer's or joint designer's specification for everything else. Note the friction condition it assumes - if it does not state one, you do not know what it means.