[Xwang]

Hi, I've a new laptop with an Intel P8400 centrino 2 duo cpu (dual core), 4gb of ram and 250 gb hdd. I work in an aerospace company and we use MGAERO on a pc similar to my laptop which is able to run a steady state simulation of flow around an airplane in some hours. Is it possible to run a similar test case with openFOAM on my laptop? Eventually, is it possible to limit open Foam to only 1 core and 3gb of ram so that I can use the laptop to navigate while it computes? Initially, I would like to run some test with a simple naca0012 rectangular wing to compare the results obtained with openFoam to the ones obtained from MGAERO. Thank you, Xwang

[Xwang]

Any idea? Xwang

[gocarts]

You can perform a steady-state simulation around an airplane, as an example (requires ~240MB of RAM for ~520k cells using simpleFoam in single precision 32-bit) try
http://www.aero.polimi.it/freecase/?...:YF-17_fighter

You can control the number of parallel processes in OpenFOAM from many down to one. You can't explicitly limit the RAM, but you can implicitly size the case by controlling the cell count and physics (e.g., turbulence model)

However, the bigger question is whether you want to run under Linux or Windows?

For Linux the default OpenFOAM distribution works fine, but if you want to run under Windows then you'll need to either patch the OpenFOAM source and try a cross-compiler or use pre-compiled binaries. For details try
http://www.symscape.com/openfoam-1-5-x-on-windows

[Xwang]

Originally Posted by gocarts

You can perform a steady-state simulation around an airplane, as an example (requires ~240MB of RAM for ~520k cells using simpleFoam in single precision 32-bit) try
http://www.aero.polimi.it/freecase/?...:YF-17_fighter

You can control the number of parallel processes in OpenFOAM from many down to one. You can't explicitly limit the RAM, but you can implicitly size the case by controlling the cell count and physics (e.g., turbulence model)

However, the bigger question is whether you want to run under Linux or Windows?

For Linux the default OpenFOAM distribution works fine, but if you want to run under Windows then you'll need to either patch the OpenFOAM source and try a cross-compiler or use pre-compiled binaries. For details try
http://www.symscape.com/openfoam-1-5-x-on-windows

I'm a linux user ans so I will install openFOAm on my kubuntu 8.10 (which I will update to 9.04 next august). Is there any specific installation howto? Xwang

[gocarts]

Originally Posted by Xwang

I'm a linux user ans so I will install openFOAm on my kubuntu 8.10 (which I will update to 9.04 next august). Is there any specific installation howto? Xwang

I've succeeded in compiling OpenFOAM on ubuntu 8.10 and 9.04, so I think you should be fine trying under kubuntu too. I didn't try compiling paraFoam which can cause problems with QT dependencies.

You are probably aware of the OpenFOAM Forum here on cfd-online, but just in case I direct you to the dedicated forum for OpenFOAM installation at:
http://www.cfd-online.com/Forums/openfoam-installation/

[Xwang]

As far as the hardware is concerning, is more important a huge amount of memory (more than 4gb) or very powerfull processor? In case I will buy a dedicated pc (under 1000€) is better an amd or an intel processor (I will use 64 bit distro ) and 8gb ram is enough for calculation around airplane with boundary layer (subsonic, transonic e supersonic compressible fstationary flows)? Xwang

[gocarts]

If your flow simulation doesn't fit into core memory and therefore needs to use swap space you will be looking at 10-100 times slower turnaround (assuming your swap resides on a disk), so I would always favor enough memory to fit your problem over a faster processor.

I haven't compared amd vs intel processors so I can't comment.

If you can get away with single precision over double precision you can cut your peak memory requirements in half - though your convergence rate might slow.

Based on the complete airplane flow simulation I mentioned earlier (assuming single precision) I think you could comfortably fit a similar density mesh with boundary layers (using wall functions and assuming twice as many cells again ~1M) under 1GB. Though I'd still go for as much memory as you can afford - no doubt you'll want room for growth as your simulation needs increase.

At some point the cost of adding more memory to a single system becomes prohibitive and it's worth considering buying a second machine with a 'reasonable' amount of memory and running the flow simulation in parallel (e.g., mpi).

Hope this helps.

[Xwang]

Ok. So I can start with my notebook (4gb memory and p8400 centrino core duo) without many problems. Probably it will be not very fast, but I can let it work during the night and when I go to work. Then in the future I can buy a dedicated pc with a more powerful processor and the possibility to scale up to 8gb. Thank you, Xwang

[albcem]

Hi,

I did perform a steady aerospace case on my 4gig memory laptop with dual-core processor with no problems. You can review the results at

http://www.caebridge.com/f-18-high-angle-attack

However, this was purely to test OpenFOAM capabilities as well as my own. I agree with one of the former comments - you are much better off with a desktop with 4+ gig memory and a 2+ cores.

Cem

[Xwang]

Originally Posted by albcem

Hi,

I did perform a steady aerospace case on my 4gig memory laptop with dual-core processor with no problems. You can review the results at

http://www.caebridge.com/f-18-high-angle-attack

However, this was purely to test OpenFOAM capabilities as well as my own. I agree with one of the former comments - you are much better off with a desktop with 4+ gig memory and a 2+ cores.

Cem

I'm in the same situation. I have to test mine and openfoam capabilities before buying a dedicated pc with an i7 quad core cpu and 6 or 12 gb ram.
Xwang