When I want to run the solver I get this error:

*** Error #329 *** Could not allocate 15739536 bytes as requested by star-cd.

I have 4G memory and 12G swap space and the model has about 7 million cells.

How can I get rid of this error?

BR,

What OS and CPU are you using?

Linux Suse 10.2 on 2 different pc.

First system: Xeon cpu and 4G ram. It is a 64bit system, but I use it as a 32bit system. I don't have 64bit software.

Second System: P4 cpu and 3G ram.

I have problem with both of them.

7 million cells can be a bit too much for 4G RAM. I was able to put about 3 millions into 2G, so the limit for 4G is may be 6..6.5 millions for turbulent incompressible flow with no additional features. It can be less if you have compressibility or heat transfer or something else.

Very true, if you are using any multi physics then you are going to eat more memory especially l2p, anything other than a two equation turbulence model will also chew memory

I am performing an automotive aerodynamic simulation. I reduce the number of mesh, but the problem there is yet. In PC version I have memory problem when I try to make mesh but there is not problem with solver. But in Linux version the problem is inverse. In other words, in LINUX I do mesh generation without any problem but with solver I need more memory. It is very strange.

Maybe the problem is that U can't use all the avaiable memory for a single process! Let try to split the problem and solve in parallel computation way! Use the metis method to split the domain, it's better!

I have only 1 cpu. Is parallel computation possible? Where can I find about metis method?

U can run a parallel computation run with a single CPU. U have to quit&save model after generating geom. and prob. files, then you click on "Run star interactively" from starlaunch panel and choose "Parallel computation on single host" in the first window of the display. Then open the advance panel an choose metis for decomposition method. Then U left blank the "Hostname" request and shift to 2 the number on the right. Try in this way and let me know if it's work. Bye

The real question is why would you want to run in parallel on 1 cpu? It will be significantly slower and take more memory than running the job serially. If its only to test something that's fine, but it will be slower.