DiPhyxDiPhyx
Workspace
The honest framing

They mostly solvethe same equations

Navier-Stokes, discretised on a mesh. What separates CFD packages is licensing, meshing, and how much of your time each one consumes.

Same physics

Conservation of mass, momentum and energy. Every package on the list solves the same system.

Meshing is the work

Most of the effort in a CFD project is geometry and mesh, not the solver run.

Licensing decides more than you expect

Per-core licences change how you can use a rented machine, which changes what a study costs.

The vocabulary

Three wordsworth getting straight

Comparison tables are unreadable without these, and they are the axes any real choice is made along.

RANS, LES, DNSHow much turbulence is modelled versus resolved. RANS for engineering, LES when the eddies matter, DNS for research.
Finite volume, element, differenceHow the domain is discretised. Finite volume dominates CFD because it conserves fluxes across cells by construction.
Pressure- or density-basedPressure-based for incompressible and low-speed flow, density-based when compressibility drives the problem.
Open source

Two you can startthis afternoon

Both are built here as workflows, so the only cost is the machine they run on.

OpenFOAM

Finite volume, RANS through DNS, a solver for nearly everything, and a steep first week.

SU2

Compressible aerodynamics and adjoint optimisation, with a far gentler configuration file.

The specialists

Palabos for lattice Boltzmann, Nektar++ for spectral element, FEniCS when you want to write the equations.

OpenFOAM is the general-purpose one: finite volume, an enormous solver library, and a case directory structure that is either elegant or infuriating depending on the day. SU2 is narrower and much cleaner to drive, built around compressible aerodynamics and, unusually, adjoint-based shape optimisation as a first-class feature rather than an add-on.

Commercial, and choosing

What a licenceactually buys

The paid packages are not better at Navier-Stokes. They are better at everything around it, and that is what should decide the choice.

MeshingYou getthe part that costs you weeks
SupportAndvalidated models, someone to call
Per coreYou paywhich caps how wide you scale

The real product is meshing, a graphical setup that catches mistakes early, validated physics models, and someone to call. For an industrial team on a deadline that is worth a great deal. For a research question where the mesh is simple and the physics is standard, it mostly buys convenience — and the licence terms often make running on rented machines awkward in a way open source never is.

Complex geometryIf meshing is the hard part, commercial meshing tools genuinely save weeks. Consider meshing there and solving elsewhere.
MultiphysicsConjugate heat transfer, fluid-structure interaction and radiation are where commercial packages still lead comfortably.
Many runs, not oneParameter sweeps and optimisation favour open source, because scaling out costs machine hours rather than licences.
ReproducibilityA text case directory under version control is easier to publish and rerun than a binary project file.
On a machine

OpenFOAM isalready built

Nothing here needs compiling. Pick a machine, and these two commands put it on there.

hub://openfoam

$ dxflow workflow create --identity openfoam hub://openfoam

$ dxflow workflow start openfoam

Start a machineWhat OpenFOAM needs
Read it, then run itPrepaid and by the hour, on a machine that is yours about a minute after you ask.
Start a machineAll notes