[maero21]

Dear CD-Adapco users,

A month ago, I started using CD-adapco STAR-CCM+ and after running some very simple models, I now try to to mesh one side (due to symmetry) of a car. Eventually I probably need around 20-30 Million polyhedral cells (lots of details).
However, when trying to volume mesh, not much happens after hitting the 6 Million cells mark.
My laptop is a decent workstation (HP 8560w) with 12GB RAM. I read somewhere that you need around 2GB RAM per 1 Million polyhedral cells, but that seems to me like a massive requirement. Then I would need a 64GB RAM workstation, which seems strange to me. Especially, since my laptop can easily mesh a 30-40 Million cell case in for instance Pointwise (tetrahedral mesh).

Is this 2GB per 1 Million cells requirement really true? Otherwise I will have to find me a new program, while I really like STAR-CCM+.

Thanks!

[ggulgulia]

hey maero

20-30 million sounds overwhelmingly massive for not only your donation, but any industrial scale problem too. You can scale down your model and. Even if there are lot of detailing, you can choose a coarser mesh size and go for a local mesh refinement at requeired locations. 2-5 million cell should be sufficient.

incase you want to go for 30 million plus cell size you will need more than 64 Gb ram. Since during simulation ram will need to store your discritized continua as well as solve the turbulence equation. You will need a hpc cluster to solve such huge mesh size case

brother. Go for a grid independence test too

[lava12005]

Hi,

20-30M polyhedral cells will absolutely need more than 12GB.
And yes, 2M poly cells ~ 1GB is quite true.
Even with trimmer mesh, 30M will use more than 12GB.
Btw, these are when you are running the simulation, not sure whether you just want to create mesh in StarCCM and running it somewhere else or you want to do everything in StarCCM. But RAM requirement for meshing is different and lower I suppose.

[me3840]

First I will disagree pretty sharply with ggulgulia, 20-30M cells, even up to 40-50M is good for an industrial scale problem. 2-5M for an external aero problem with a complicated geometry is woefully insufficient.

But 2GB of memory per million cells is about right.

Remember about 6 tetrahedrons need to be combined to make 1 polyhedron. So while you can generate a grid in Pointwise that's 30-40 million tets, to make a similar mesh with polyhedrons you would only need 5 million cells or so. If you want to generate 20M polys, you need about 120M tets.