What Actually Sets the “W” in 0W-20: The Two Cold Tests Behind a Winter Grade

Last verified: 2026-08-04

Suppose a delivery van, left overnight in an unheated yard, starts on the first crank the next cold morning. The engine sounds completely normal. Somewhere inside the block, though, the oil pump may already be pulling nothing but air from the pickup — not because the engine failed to start, but because the oil failed to move once it was needed. Starting and oiling are two different problems. The "W" in a grade like 0W-20 is not one number that certifies both of them. It is the record of two separate cold-laboratory bench tests, and the second one measures a failure the first test cannot physically see.

Every source that touches this question — including a fair amount of technical writing already online — starts and stops at one line. SAE J300, in its current edition (revised May 2024, numbered J300_202405), states in its own scope that it "defines the limits for a classification of engine lubricating oils in rheological terms only," and that "other oil characteristics are not considered or included." That sentence is accurate, and for most of what gets published about it, that is also where the explanation ends. It confirms the W is a laboratory classification, not a weather forecast. It does not say what either of the two machines behind that classification actually does, why they exist to catch two different failures, or why one of them can report something the other physically cannot.

Why does a winter grade need two different lab tests?

The reason traces to a documented failure, not a formulating preference. A large number of engine oil pumpability failures occurred in the field during the winter of 1980–81, and the pumpability literature divides them into two populations: oils that failed after a single overnight exposure to low temperature, and oils that needed several days of exposure before problems developed.

That split is the whole argument for a slow test. Cranking happens once, at the very start of the event. The second population's failure took days of cold soak to build, and no quick chill reproduces it.

The instrument in force at the time did not catch those oils. SAE technical paper 831714 (Smith, 1983) records that the mini-rotary viscometer, adopted by SAE in 1980 to predict engine oil pumpability, "was unable to fail certain commercial oils that caused engine problems in the winter of 1980-81" — and it introduces a revised, more effective cooling cycle for the instrument as the fix. The temperature-profile approach that came out of that period is the direct ancestor of the pumping test used today, ASTM D4684.

What does the cold-cranking simulator actually measure?

ASTM D5293 answers the first question: can the starter turn this engine over at all? The method spins a rotor inside a close-clearance stator at very high shear stress — on the order of 50,000 to 100,000 Pa, by ASTM's own description — a reasonable proxy for what a starter motor does to the oil film in a bearing clearance during cranking. The result is reported as an apparent viscosity, in mPa·s, typically somewhere in a range of roughly 900 to 25,000 mPa·s depending on the oil and the test temperature. ASTM states plainly that the result correlates with actual low-temperature engine cranking.

On a data sheet, this shows up as a single CCS apparent viscosity figure quoted in mPa·s at one stated test temperature. Read the temperature sitting next to that number — the number alone means nothing without it.

D5293 answers whether the starter can turn the engine over. It says nothing about what happens in the seconds after, once the crankshaft is already spinning and the pump has to move that same oil up through the pickup, past the filter, and out to the bearings and the valvetrain. That is a separate mechanical event, and it needed its own test.

Can an engine crank normally and still starve for oil?

Yes — and this is the failure D5293 cannot see. Cooled far enough, and given time, the oil doesn't just get thicker. It can turn to something closer to a jelly than a liquid. The paraffinic wax naturally dissolved in the oil crystallises out below a certain temperature and grows into an interlocking, three-dimensional network that traps the liquid oil inside it. The bulk fluid picks up a yield stress — a threshold force that has to be exceeded before the material flows as a liquid at all. A pump built to draw a liquid cannot draw a gel. It draws air from the pickup instead, a failure mode the wider industry calls "air binding" — though that is general usage, not a term defined inside ASTM D4684's own text.

D4684 is built to catch exactly that. It applies a shear stress of about 525 Pa at a shear rate of roughly 0.4 to 15 s⁻¹ — a far gentler regime than D5293's cranking-level shear, because it is simulating suction through a pickup tube, not a starter turning a crankshaft. ASTM's own scope describes the failure directly: the oil "forming a gel structure that results in either excessive yield stress or viscosity," and states the method has "predicted as failures the known engine oils that have failed in the field because of lack of oil pumpability."

The result, when it is reported at all, is two separate numbers at a stated low temperature — not one. D4684's own title names both: apparent viscosity, and yield stress. A data sheet that quotes only a viscosity figure from this test has published half the result.

Why does the cooling path change the result?

Because the process building that wax network is kinetic, not an equilibrium state the oil simply settles into. Cooled slowly, through the temperature window where wax first crystallises, the crystals have time to grow large, well-ordered, and heavily interlocked. Cooled quickly to the same final temperature, the wax still crystallises — it has no choice, thermodynamically — but the crystals that form are smaller, more numerous, and far less connected to one another.

The same oil, cooled to the same final temperature, can therefore show a very different yield stress depending only on how slowly it got there.

That is why D4684's cooling protocol is not a quick chill. It runs a slow, controlled cool through the crystallisation range, followed by a faster final step down to the actual test temperature, with the total controlled cooling exceeding 45 hours before a reading is taken — test temperatures span roughly −10 °C to −40 °C. A rushed cooling profile would understate the real risk; an overnight cold soak in a real vehicle sits closer to the slow end of that spectrum than to a quick lab quench.

Pour point depressants exist to interrupt exactly this network, and they do it without stopping the wax from crystallising at all. They are comb-shaped polymers — poly(meth)acrylates, ethylene-vinyl-acetate copolymers, and related chemistries — built with a long non-polar segment that co-crystallises alongside the natural wax, and a bulkier, polar segment that will not fit into the growing lattice. The wax still forms crystals. It just cannot link them into one continuous structure, because the polymer keeps interrupting the growth.

This is also why pour point itself (ASTM D97, or its newer robotic tilt alternative, ASTM D6892, current as of November 2025 and reporting results at 3 °C intervals) is a weak predictor of pumping performance. Pour point measures whether the oil is observed to stop flowing under a specific rapid-cool protocol — a quench, not a soak. It does not measure yield stress built up over 45-plus hours the way D4684 does, and wax composition — the mix of normal and branched paraffins present — matters to the outcome as much as the raw quantity of wax does. A favorable pour point does not, by itself, guarantee a favorable D4684 result.

Is there a third test, and what does it add?

There is: ASTM D5133, the Scanning Brookfield technique, introduced originally for the ILSAC GF-2 passenger-car specification. Instead of one reading at one final temperature, it continuously scans viscosity while cooling the sample from about −5 °C down to −40 °C at a controlled rate of 1 °C per hour, and reports a Gelation Index — the maximum rate at which viscosity increases across that scan — along with the temperature at which that maximum occurs.

The distinction D5133 draws is useful on its own terms: flow-limited behaviour is associated with the oil's viscosity, while air-binding, gel-type behaviour is associated with gelation — two different mechanisms, scored as two different numbers rather than inferred from one.

A related, older method is worth naming so it is not confused with D4684: ASTM D3829, the borderline pumping temperature test, run on the same family of instrumentation, using a 16-hour cool from 0 °C down to −40 °C to yield a single borderline pumping temperature. It remains an active, current method — it was not withdrawn — and it is used in some other specifications. SAE J300 today uses D4684 for its pumping limit, not D3829; the two methods are not interchangeable, and a data sheet that cites one is not implicitly reporting the other.

Does a colder W automatically mean a safer choice?

No — and this is where base oil chemistry and additive treat rate start to matter more than the grade name on the drum.

API Group IV base stock — polyalphaolefin, built by oligomerising alpha-olefins rather than refined out of wax-bearing crude fractions — contains no paraffinic wax by construction. It does not show the gelation mechanism described above at all; whatever cold-temperature behaviour it has is ordinary viscous thickening, not a crystallising network. Directionally, Group I stocks retain the most natural wax and lean hardest on pour point depressant chemistry to control it; Group II and Group III stocks are more severely processed and generally need less.

Viscosity index improvers are implicated too, and not only at the hot end of the range where multigrade formulation usually gets discussed. These polymers coil tightly when cold, contributing comparatively little to viscosity at a cold start, and uncoil as the oil warms — that differential behaviour is what makes a multigrade oil multigrade in the first place. The chemistry choice carries a real trade-off: olefin copolymer viscosity modifiers are highly efficient thickeners at low cost but generally show weaker shear stability, while polymethacrylate modifiers behave better at low temperature and hold their shear stability longer, at a lower thickening efficiency per unit of mass and a higher cost.

There is also a ceiling working against the impulse to chase the coldest possible number. Reaching for a lighter, lower-viscosity base-oil fraction to hit an aggressive cold-flow target runs straight into Noack volatility (ASTM D5800): lighter fractions have a lower average boiling range and are inherently more prone to evaporating away at operating temperature. A formulation optimised purely for cold pumpability can end up fighting its own volatility limit.

There is no single "right" W, and that is not a dodge — it names three variables that actually decide it: the equipment's real cold-duty cycle (how cold, and how long a soak typically sits between starts), the base-oil group behind the formulation, and the additive treat rate — pour point depressant dose and VI-improver chemistry — supporting the claim. The per-grade CCS and MRV limits and their test temperatures live inside SAE J300's own table, and that table is paywalled and has changed across editions; a data sheet quoting a specific grade should be read against the edition it was actually tested to, not against a number recalled from memory. One structural pattern is worth knowing even so: the pumping test temperature assigned to a given W grade runs colder than the cranking test temperature assigned to the same grade — but the exact figures belong to the current standard, not to a summary.

What should you request in writing before you specify a winter grade?

Start with the two numbers this whole piece has been building toward, and be specific about what "the number" actually means on each test.

None of this is a reason to distrust a 0W-20 label. It is a reason to ask what stands behind it. The how to read a lubricant technical data sheet guide walks through exactly where figures like these sit on a real data sheet, line by line; for how the W fits into the wider engine-oil grade structure it belongs to, see the petrol engine oil specifications guide. Where a genuine cold-climate comparison is the real sourcing question — one supplier's D4684 result specified for one operating region against another's for a colder one — put both figures, with their test temperatures, directly into the request for quotation sent to suppliers listed under petrol engine oils, rather than relying on the grade name alone to carry that distinction.

Frequently asked questions