Learning Forum

[eugenzilio]

Hi. I'm running a steady RANS simulation with say 500k cells. I'm using a laptop with the following specs:

CPU - Processor Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz, 2592 Mhz, 6 Core(s), 6 Logical Processor(s)

RAM - 16 GB

I disabled hyperthreading and I am now running simulations using 1, 2, 3 cores and so on and comparing the speed. I noticed that the speed does not scale beyond 5 cores, so I started looking into parallel processing in Fluent to see what I could find. As there could be numerous hardware and software related reasons, the Fluent User Guide suggests using Parallel - Check to troubleshoot. I executed the Check command I got the following message:

WARNING! Memory utilization is very high or imbalanced.

I executed Parallel - CPU Info - Memory Usage and got the following information:

----------------------------------------------
| Virtual Mem Usage (GB)|
ID | Current | Peak | Page Faults
----------------------------------------------
host | 0.167267 | 0.170776 | 5.442e+04
n0 | 0.643375 | 0.688896 | 8.343e+06
----------------------------------------------
Total | 0.810642 | 0.859673 | 8.397e+06
----------------------------------------------

In my 'professional' opinion, the memory usag does not seem very high :). Interestingly, I get this error even when I'm running on 1 core. It seems that I don't understand something so I'm hoping for some help.

EDIT: The warning does not show up for 3 cores but it does for 1 core and 6 cores.

[Rob]

Depending on physics you'll lose efficiency somewhere between about 50-100k cells per core. This can also be dependent on the model shape and chip interconnect speed.

[eugenzilio]

thank you for your answer. Could you provide a source for your statement about the cell to core ratio? I believe it should be somewhere in the Fluent Guide but I haven't found it yet.

[Rob]

I don't think it's in the manuals, it's from the parallel speed up graphs that are posted when we release a new version. Given the number of possible model combinations it's a rule of thumb rather than a precise value.

[eugenzilio]

Makes sense, now that you say it :). Thank you for your help!