[CaptainCombo]

Hi all,


I am working on simulation of forced convection from heated blocks in a channel. SST k-omega Intermittency Model is employed for tubulence modeling and numer of cells(hexa) is around 20M.


The simulation is runing on eight E5-2650 v4(hyperthreading is disabled) processors and the system has 64Gb DDR4 2400Mhz quad channel memory.



1. The simulation has advanced only 0.55 seconds in two days. Up until now I have used Fluent with desktop computers for simpler simulations therefore I can't make performance comparsations with my old simulations. What I want to learn is whether this performance is normal or not? I am aware that advancement speed of simulation highly depend on problem but still I want to know opinion of experienced users. Because there is no way to compelete simulation at this speed.


2. There is another issue that make me suspicious about hardware. When I test same problem with 12, 24, 36, 48, ..... 96 cores, I see that 84 cores give best performance. Why it is not possible to get better performance with 96 cores?