Hi All,

I have a question about computer resource.

If a CFD model with more than four million cells is to be solved for turbulent flow by Fluent6.2. The results should be within only several days. How many CPU, and how fast GH and how large GB are min requred. Which computer system is best fit to this case? Windows XP or Unix system?

Your comments would be very grateful.

Thanks,

Mark

Unix machine will be the best. Setup the problem and run it in batch mode without loading GUI.

-H

Strongly agree with Harish.

Hi Harish,

If in terms of CPU, speed GH and GB what is min requrement for that model and which kind of Unix model you suggest will be best fit for that case?

Thanks,

Mark

best would be a 3 ghz processor with 2 gb of ram . Any variant of linux or unix is fine .

-H

I believe that you will need about 4 Gb of ram. Try to use dual cpu linux box, so that you can run parallel fluent with 2 cpu's. Anything above 2.8 Gz should be fine. Running time will depend on which turbulence model you use and whether steady or unsteady simulation is required.

i have four 64 bit linux machines with 4gb ram each, for the cases around 3.5 million cells i can run it on double machines, its okey, but for cases with cells more than 4 million i have to move to three machines, two machine system would also be fine but it becomes too slow.

in all two machines would eb okey, with one machine you wont be able to run your case comfortably, it becomes too slow.

Thanks all for your comments. It is very useful information.

The cases with more than 4 millions cells (for example unsteady turbulent flow with VOF model) are very often in industry. Usually they are requred to be solved in couple days. It needs computer with fast and large memory.

Let us discuss which computer will be best for that case? 32bit WindowXP or Unix and linx 64bit system with 3GH speed or higher (what is the fastest speed so far ?)and 10 GB RAM? What is different for 32bit and 64bit and 72bit system for that case?

Which model of computer (WindowXP, WindowNT, Unix or Linx ???)will have these functions? Any suggestions will be welcome.

Thanks in advance,

Mark

I posted a topic not so long ago regarding AMD Opteron. From answers it is clear to me that Windows can't handle it. Mostly because of bad memory management and memory limitation for a single process.

With such a large no. of nodes I think you should talk in terms of flops rather than GHZ.

Saad

Presently I am running a case with 4.1 million cells. I m using 8*2GB of RAM. Still it is slow as the model is complicated. Always, unix/linus is better than windows. Its faster. Also most of the times, I m facing problem of node coming out i.e. communication problem on Windows. Better to use only one(or minimum) system/s with higher RAM.