The comparison only means something if the alloy and temper match

Most of the confusion in this debate comes from comparing the wrong things. A 6061-T6 extrusion is a heat-treated wrought alloy, and a heat-treated A356 casting is a different alloy system that has also been heat-treated. Comparing a T6 extrusion to an as-cast part is not a comparison of processes; it is a comparison of a heat-treated part with an untreated one, and the heat treatment is doing most of the work.

For reference, 6061-T6 is commonly quoted at an ultimate tensile strength of about 310 MPa and a 0.2 per cent proof stress of about 276 MPa, with good elongation. A well-made, heat-treated casting of a comparable alloy lands in the same order of ultimate strength but typically lower in proof stress and markedly lower in elongation, because the casting still contains the defects that the extrusion process squeezed out.

So the first rule is to compare like with like: same alloy family, same temper, same heat treatment, and the same section. Then the remaining differences are genuinely process differences, and they are the ones worth designing around.

Extrusion gives grain flow, which makes strength directional

In extrusion, a heated billet is pushed through a die and the metal flows. The grains deform and elongate in the direction of flow, so the finished section has a fibrous grain structure running along its length. That grain flow is why an extruded part is strong in the extrusion direction: a crack that tries to run across the section has to cross the grain boundaries rather than follow them.

The same structure is why the properties are directional. Across the extrusion direction the grain flow gives less benefit, and a load that tries to split the section, or to open a slot, is working against a different material than a load that bends the section along its length. On a T-slot frame the practical case is a slot lip loaded by a bolt: the load is not along the extrusion axis, and the strength there is governed by the lip geometry and the local material rather than by the longitudinal properties quoted in a datasheet.

There is a second microstructural feature worth knowing. The outer skin of an extrusion cools and deforms differently from the core, producing a surface layer with a finer or recrystallised structure. That layer is usually beneficial for surface finish and corrosion, but it is different material from the core, and a machining operation that removes it changes the surface behaviour of the part.

Hollow extrusions have a weld seam, and it is a discontinuity

A solid extrusion is pushed through a simple die. A hollow extrusion needs a die with a mandrel or a bridge, and the metal has to flow around the supports and rejoin on the other side. Where it rejoins, it forms a weld seam running along the length of the profile. A second type, the charge weld, is a transverse joint formed between successive billets in a press run.

These seams are not necessarily defects, and a well-run press produces a seam that is metallurgically sound. But the seam is a place where the structure and the properties differ from the bulk material, and it is a preferred path for certain failure modes. In a pressure-tight application a seam is the first place a leak appears. In a heavily loaded hollow section, the seam is a place to check rather than assume.

The practical consequences are modest but real. A hollow profile that has to be pressure-tight needs a seam-quality requirement rather than a general one. A hollow profile that is machined through should have the seam considered in the machining plan. And a buyer comparing a solid and a hollow section of the same envelope should know that the hollow one is not simply a lighter version of the same material.

Casting trades porosity for shape and isotropy

Casting fills a mould with liquid metal, which means the shape can be complex in three dimensions: ribs, bosses, thick-to-thin transitions and internal passages that no constant-section extrusion can produce. For a part that has to be a housing, a manifold or a bracket with a curved load path, casting is often the only sensible route.

The cost of that freedom is a different defect population. Liquid metal shrinks as it solidifies, and the shrinkage has to be fed; where it is not, the part contains shrinkage porosity. Gas dissolved in the melt can be trapped as gas porosity. Oxide films can be folded into the metal during filling. All of these are internal discontinuities, and all of them concentrate stress. Good foundry practice controls them, but it does not eliminate them, and a casting is therefore specified with a quality level and inspected for internal soundness rather than assumed to be uniform material.

The compensating advantage is isotropy. Because the grains are equiaxed, a casting has essentially the same properties in every direction, so a complex load path does not have to be aligned with a grain flow. That is why castings dominate for housings and manifolds, and why the choice between the two processes is often really a choice between a simple load path and a complex one.

Fatigue is where the two processes separate most clearly

Aluminium has no fatigue limit: unlike steel, there is no stress range below which it can be cycled indefinitely, so the S-N curve keeps falling. That makes every aluminium part a fatigue question, and it makes the defect population decisive, because fatigue cracks start at discontinuities.

An extrusion starts with fewer internal discontinuities than a casting, so for the same nominal stress it usually has the better fatigue performance. But an extrusion has its own initiation sites: the weld seam in a hollow profile, a sharp die line, and any machining mark. A casting starts with porosity and shrinkage, and each pore is a potential initiation site, which is why a casting is more sensitive to the size and distribution of its defects than an extrusion is.

The design consequence is that a fatigue-sensitive part should be specified with its discontinuities in mind. For an extrusion, that means knowing where the seam is and keeping it out of the highest-stress region. For a casting, that means specifying a soundness level and verifying it, and avoiding designs that put the highest stress where the porosity is most likely to be.

Choosing between them on volume, shape and load path

Three questions decide it, and they are usually easy to answer.

The first is shape. If the part is a constant-section member — a beam, a post, a rail, a frame member, a heat sink — extrusion wins, because the die is amortised over length and the section can be optimised for bending. If the part is a three-dimensional housing with internal features, casting wins, because machining an extrusion into that shape wastes material and time.

The second is volume. Extrusion tooling is inexpensive relative to its output, so extrusion is economic from short runs upwards, and a custom die is often justified for a single project. Casting tooling is more expensive, so casting needs volume to amortise it, and a low-volume casting is usually a machined extrusion or a machined billet instead.

The third is the load path. If the load runs along the member, extrusion gives the grain flow for free. If the load path is complex, or if the part has to carry load in several directions at once, the isotropy of a casting is worth paying for. And if the part is fatigue-critical, the defect population matters more than the process name, so specify the soundness or the seam quality and inspect it.

What to send when the process is still open

If the choice between an extrusion and a casting is still open, the enquiry should carry the load case, the load direction relative to the part, the number of parts, the fatigue requirement if any, the tolerances that matter, and the surface finish. With those, a supplier can say honestly which process fits, and whether a hybrid — an extruded section with machined features, or a casting with machined interfaces — is the better answer.

What does not work is to ask for the stronger material. Strength is directional, it depends on the alloy and temper, and on a real part it is usually the joints and the discontinuities that decide the outcome rather than the number in the datasheet.