CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Hardware (http://www.cfd-online.com/Forums/hardware/)
-   -   Estimation for memory usage by ICEM (http://www.cfd-online.com/Forums/hardware/92112-estimation-memory-usage-icem.html)

ftab September 2, 2011 12:06

Estimation for memory usage by ICEM
 
Hi Everybody,
I am about to buy a new computer and my bottle neck of work is mesh generation for a case with 15,000 K cells.
Do you have any rough estimation of ram usage of ICEM for mesh generation?
Something like xMB of ram per 1k tetrahedral/hexahedral/ cells.

This will help me a lot.

Cheers,
Ftab

Tony12 September 12, 2011 19:16

I just checked my computer while running the solver for the tutorial Supersonic Missile (CFD Tutoritals > Cart3D > Supersonic Missile). Here is the information outputted when I first loaded the completed project:

10 densities
11 parts
Element types :
HEXA_8 : 402216
TRI_3 : 70154
Element parts :
VOLCELL : 377718
Total elements : 472370
Total nodes : 484311


The solver was using 280MB of memory, and ICEM was using maybe 170MB. So 450MB total. My computer is running 32-bit, Windows 7, dual core processor @ 1.79GHz (each core), 3.49 GB memory useable. I hope this helps. The simulation took a little time, approximately 2 hrs, to complete. But it wasn't too bad. So any system faster than mine would be enough. I hope this helps!

abdul099 September 12, 2011 19:39

Quote:

Originally Posted by ftab (Post 322678)
my bottle neck of work is mesh generation for a case with 15,000 K cells.

Quote:

Originally Posted by Tony12 (Post 323839)
VOLCELL : 377718
The solver was using 280MB of memory, and ICEM was using maybe 170MB. So 450MB total.
any system faster than mine would be enough.

Tony, nearly EVERY machine on earth is faster than yours, but way not sufficient for a case with 15,000k cells. That's about 40 times more than your case!

ftab, I'm sorry I can't help you with any specific information as I don't use ICEM. But why don't you try it on your own? Create some meshes and monitor memory usage and check the mesh size after meshing is finished...

Tony12 September 12, 2011 21:40

Quote:

Originally Posted by abdul099 (Post 323845)
Tony, nearly EVERY machine on earth is faster than yours, but way not sufficient for a case with 15,000k cells. That's about 40 times more than your case!


Yeah this thing is a few years old, but even a few years ago it wasn't much haha. Plus my integrated video card doesn't make things any better :/

Either way it still runs the solver without too much of a problem unless someone is looking for solutions much quicker.

abdul099 September 18, 2011 18:39

I agree, it doesn't need the fastest machine on earth to get a solution. Especially when memory is not an issue. But to mesh and solve quite big cases, there is something more needed. A faster CPU or more cores are nice, but enough memory is essential!

In my last post, I just referred to your sentence, any machine faster than yours would be sufficient. That's not true, as a machine with lets say a core2quad and 2GB memory will also not be sufficient (just an estimation from the numbers given by you). It's dangerous for other people to make any statements which could be misunderstood. :eek:

ftab September 19, 2011 04:32

Quote:

Originally Posted by abdul099 (Post 323845)
But why don't you try it on your own? Create some meshes and monitor memory usage and check the mesh size after meshing is finished...

This is exactly what I have done, but this only gives me the upper limit of my own computer not a suggestion to buy a better one.
Any way, I tried and solved it and I would say 16 Mb of memory can handle mesh up to 50M elements. My supervisor was also generous to buy a 48Mb Ram pc for me. So the problem is completely solved:D


All times are GMT -4. The time now is 20:54.