Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

question about the memory

Please login with a confirmed email address before reporting spam

Hi to everyone,

I would like to ask a question about the memory of comsol 4.3.
When it formulates the mathematical problem comsol needs to construct some matrices which will result to one or two matrices which in turn will be soved. So the memory, that it needs for the assembly of the matrices, is provided by the RAM or from the hard disk of the pc?
My problem is that when I try to solve my structure with fine mesh I do not have any error but when I try to solve with extra fine mesh then I have the following error : Out of memory during assembly.
But the RAM memory of the server that I use is 64Gb, so it seems to me very strange to have problem on the RAM memory. I beleive that the matrices do not have any problem because if they had, I would not be able to solve it for fine mesh.
Thanks in advance.

5 Replies Last Post 2012年11月30日 GMT-5 02:03
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年11月23日 GMT-5 09:43
Hi

indeed that is quite some RAM, but depending on how many DoF you have it might not be enough. Have you checked what kind of RAM you need ? there is a KB knowledge base document discussing DoFs and approx RAM and mesh types

--
Good luck
Ivar
Hi indeed that is quite some RAM, but depending on how many DoF you have it might not be enough. Have you checked what kind of RAM you need ? there is a KB knowledge base document discussing DoFs and approx RAM and mesh types -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年11月23日 GMT-5 10:50
I am impressed by your immediate response. Thank you very much.
My structure is 3D and has 663000 elements.
I have 7 different dielectric layers which have same length and width but very different height. For exaample three of them are X nm , 3 of them 4*X nm and 1 layer of 30*X nm. It is obvious that there is a big ratio between the height that is why I use finer and extra fine mesh, otherwise I have some instabilities in the solver. I have used swept meshing for each boundary-domain. Furthermore, the whole structure is excited by two lumped ports from 1Ghz to 10Ghz
I suppose that I will find this knowledge base in manual. If I am wrong please tell me where to find it.
Thank you again.
I am impressed by your immediate response. Thank you very much. My structure is 3D and has 663000 elements. I have 7 different dielectric layers which have same length and width but very different height. For exaample three of them are X nm , 3 of them 4*X nm and 1 layer of 30*X nm. It is obvious that there is a big ratio between the height that is why I use finer and extra fine mesh, otherwise I have some instabilities in the solver. I have used swept meshing for each boundary-domain. Furthermore, the whole structure is excited by two lumped ports from 1Ghz to 10Ghz I suppose that I will find this knowledge base in manual. If I am wrong please tell me where to find it. Thank you again.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年11月23日 GMT-5 13:41
Hi
if I replied rapidly it was just fair luck, I was waiting 30 sec for Comsol to finsih a solver sequence, so I took a visit to the Forum ;)

Now 600kDoF is not that much, why should that give RAM issues when you have so much ?

OK if you have fine layers the meshing is tricky, and at high RF frequencies you need to have also many elements per wavelength * index of refraction.
Have you considered if it's possible to transform some of the thin layers in "surface physics" ? sometimes this helps, else you would need N elements in the thickness, at least enough to solve your depedent variables, and their slopes

--
Good luck
Ivar
Hi if I replied rapidly it was just fair luck, I was waiting 30 sec for Comsol to finsih a solver sequence, so I took a visit to the Forum ;) Now 600kDoF is not that much, why should that give RAM issues when you have so much ? OK if you have fine layers the meshing is tricky, and at high RF frequencies you need to have also many elements per wavelength * index of refraction. Have you considered if it's possible to transform some of the thin layers in "surface physics" ? sometimes this helps, else you would need N elements in the thickness, at least enough to solve your depedent variables, and their slopes -- Good luck Ivar

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年11月29日 GMT-5 12:54
Thank you very much for your response. The problem was that I had a lot of elements in swept distributiion. So this problem has been solved. I would like to ask you another question if it is possible. I would like to calculate the impedance of this multilayer, what would be the appropriate boundary conditions?
Thanks in advance.
Thank you very much for your response. The problem was that I had a lot of elements in swept distributiion. So this problem has been solved. I would like to ask you another question if it is possible. I would like to calculate the impedance of this multilayer, what would be the appropriate boundary conditions? Thanks in advance.

Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 2012年11月30日 GMT-5 02:03
Hi
for quick impedance calculations, you should study the lumped port / terminal BCs in the doc and compare these to the electric potential or electric current BC (which are for field variables BC values, and not scalars)

--
Good luck
Ivar
Hi for quick impedance calculations, you should study the lumped port / terminal BCs in the doc and compare these to the electric potential or electric current BC (which are for field variables BC values, and not scalars) -- Good luck Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.