i have a 3 gb ram windows machine, i am on XP paltform, i was wondering what is the maximum size of case some body has run on this kind of RAM (assume we have only one machine to run).

or can i run 3.5 million cells case on this 3gb ram machine.

On 32bit XP, 2Gb is your limit sadly!

What mesh type are you using? Hex or Tet mesh?

James is right. Is your machine 32bit or 64bit?

Hi,

Assuming your machine is 32 bit and your simulation is single phase flow, you will be able to run about 700k nodes as a tet or tet/prism mesh, or about 1 million nodes as a hex mesh. To go bigger you need either 64 bit or to run the job in parallel. These are ball park figures only, so you might be a bit higher or lower.

Regards, Glenn Horrocks

i put this question because, i have been trying to write a segregated solver, and with three levels of multigrid, i see that my code takes around 525 MB per million cells, and since i have spent around 10 days just to add this new memory model to my code, that allocates memory very wisely, and no matter how i try i am not able to bring the memory requirement per million cell less than this.

well i read somewhere (on ansys web page ) that for 1 million cells CFX takes around 400 mb. and since CFX is coupled solver, i wonder how they can manage in such a small memory limitations. now since i do not have access to CFX, i thought lets ask users what they feel about it.

and as James said, 2GB is limit on XP, by that i guess CFX can at least take 4 x 400 mb and hence a CFX user on a 3GB ram machine like mine can work, with at least a case with 3.5 million cells, (that i am unable to allocate in my solver on winXP), the maximum size i am able to allocate in my solver is 3.1 million cells (all hex), and after that my solver try to take more than 2 gb so, xp gives me out of memory error(yes it is 32 bit machine).

we shall assume that over case is single phase with one 2 equation turbulence model on (so it needs to allocate memory for it).

Glenn 700K nodes means how much cells roughly???

Hi,

Be careful when you compare different codes. CFX is a node based solver with some finite element aspects to it (for instance integration points at different locations to the nodes). Your segregated solver may store variables at the nodes, elements or a combination of both.

Regards, Glenn

Actually i don't wish to compare, but CFX claim that 1 million cells in 400mb was something i was finding it hard to stomach, i wished if some users will also say that yes we also feel that we can run the case with 1 million cells in 400mb.

as far as comparision is concerned, i shall compare my solver with fluent , since it reads fluent's file, and in that case my solver (at least till now) beats fluents by a good margin (my code is able to read and allocate for 3 million cells (hex) my windows machine, where as fluent fails to head even 2.5 million (hex) on same machine, but then i shall consider this that they allocate space for visualisation, post process etc, and i give user an option not to allocate all that).