Workstation Performance Optimization
 

Hi all,

I have HP Z640 workstation with DUAL E5-2650 v3 processor. It have 10 cores and 2.3 Ghz for each processor.

PC also consist 8 number of 8GB DDR4-2133 ECC Registered RAMs for each RAM slot and 64 GB totally.

It also have 250 GB SSD and 2 number of 2 TB Harddisk.

I am going to be use this machine for ICE simulation with ANSYS Forte. I'm asking that what types of setting changes should I made to get best simulation solution time.

As I know, hyperthreading is useless for CFD so I disabled it. But I want to learn how many core should I use for simulation? (20,16,12,10, ?) My friend told that don't use all cores for simulation, ıt extends the simulation time, use max 16 core. Is it correct?

I also have to installed sofwtare to HD not SSD because of lack of disk space. Is it make sense on the simulation time?

Thank you very much in advance.

Edit: I run a simple case with total 20 cores and 10 cores useage. The results are very interesting. 10 core takes 17 minutes but 20 core takes 47 minutes. Why is there a huge difference?

Are there anybody for help me?

There could be many reasons 20 cores was slower than 10. All problems are different though, some scale better than others, as do different codes.
- A small problem that does not distribute well will likely get slower as you are splitting up the problem too much, and adding overhead which outweighs the additional cores. Or it is memory bandwidth limited, not CPU limited, and adding extra cores again, just adds overhead. this is very likely with 20 cores and only 8 memory channels.
It doesn't matter where you install the software, other than program loading time. The working directory will affect your simulation time more if the program is using disk space for a swap drive (if you are solving out of core) and the SSD would help here. I'm not familiar with ANSYS Forte and if is uses disks for swap space?