CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   4 core & 8 core: same 64bit XP, Xeon, x5570, 2.93GHZ (http://www.cfd-online.com/Forums/cfx/67356-4-core-8-core-same-64bit-xp-xeon-x5570-2-93ghz.html)

leeruoyun August 12, 2009 16:09

4 core & 8 core: same 64bit XP, Xeon, x5570, 2.93GHZ
 
for same *.def file, run both on 4 core & 8 core workstation (other config are same: 64bit XP, Xeon, x5570, 2.93GHZ)

see cpu info here: http://picasaweb.google.com/leeruoyu...56084054998514

4 core is converged within 300 iterations, while 8 core not converged within 300

why ? CPU bothers ?

thanks for advice !

version: CFX12 64 bit

ghorrocks August 13, 2009 00:06

It is unusual for adding extra processes to affect the convergence of the simulation. What type of simulation is it?

Glenn Horrocks

leeruoyun August 14, 2009 03:58

thank you !

it is really strange. any random initilization for CFX ? e.g., how solver to kick off the flow with the prescribed turbulent intensity ? In this case, we would always get different results for different run (i do not believe this as when i run 2 time in the same pc, results are same), puzzled ........ slower or faster we should always get the same thing ah ...

ghorrocks August 16, 2009 09:24

Hi,

CFX only uses random numbers (to my knowledge) in modelling lagrangian particles with turbulence and in generating inlet/initial conditions for LES models. I do not know of any other random numbers used in CFX - but I am not a developer so I am only guessing.

You still have not described what you are doing.

Glenn

leeruoyun August 19, 2009 01:47

hi i am doing porous media (seepage flow) problem... i will test another case soon (free surface problem) ...


All times are GMT -4. The time now is 14:19.