DiPhyxDiPhyx
Workspace
The question behind the question

Nobody is askingwhich solves it better

Both solve Navier-Stokes. The argument is about meshing, support, licence terms and who absorbs the risk when a result is wrong.

The physics is shared

Finite volume, the same turbulence models, the same discretisation schemes. Validated against the same cases.

Meshing is the work

Most of a CFD project is geometry preparation and mesh generation, and this is where packages differ most.

Somebody carries the risk

In a regulated or contractual setting, a supplier who stands behind the result is part of what is being bought.

What you pay for

Four things,and none of them is the solver

A commercial licence buys real value. It is worth being precise about what that value is.

Meshing that worksRobust automatic meshing on dirty CAD is the strongest single argument for a commercial package.
A setup that catches mistakesA graphical workflow that refuses an inconsistent boundary condition saves days you would otherwise lose.
Somebody to callSupport with a contract behind it, and validated models for specific industries. Real, and worth money.
Multiphysics, integratedFluid-structure interaction, conjugate heat transfer and radiation, working together without a coupling script.
What open source gives back

Things a licencecannot buy

These are not cost savings. They are capabilities that a closed package structurally cannot offer.

You can read the solver

When a result surprises you, the discretisation is in front of you rather than behind a support ticket.

Scale without a licence

No per-core fee, so how wide you run is a question of machine hours and nothing else.

Reproducibility

A text case directory under version control. Publishable, rerunnable, and readable in ten years.

You can change it

New boundary conditions and custom solvers are a supported activity, not a feature request.

The scaling one is the least appreciated. A per-core licence means adding cores adds cost twice, so there is a ceiling on how wide you run that has nothing to do with the physics. With OpenFOAM or SU2 the only limit is machine hours, which makes parameter sweeps and optimisation studies a completely different proposition.

The real costs

Free is notfree

Open source moves the cost from a licence line to a salary line, and that trade is not always favourable.

LicenceCommercialand a cap on how wide you run
Your timeOpen sourcethe first month is genuinely hard
Machine hoursBoththe run costs the same either way

OpenFOAM's first month is genuinely hard: the case directory structure, the dictionary syntax, and meshing with snappyHexMesh all take real time to learn. If nobody in the group knows it, that time is the true price, and for a team with one deadline and complex geometry a commercial licence is often the cheaper answer honestly assessed.

One deadline, complex CADCommercial. The meshing alone saves more time than the licence costs, on a real deadline.
Many runs, or optimisationOpen source. Per-core licensing makes wide parameter studies structurally expensive.
Research you will publishOpen source. Reviewers can rerun your case, which they cannot do with a binary project file.
Often, bothMesh in a commercial tool, export, then solve in OpenFOAM. A common and legitimate arrangement.
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