Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Does increasing the number of processors speed up the "wrf" simulations while using MPI run ??

Tanmoy

New member
Hello,

I've started using WRF over the last year. From NCAR tutorials and WRF user guide, I came to know that using more than one processor would improve my model performance and also optimize the total model run time while using mpirun. So, accordingly, I've followed the guide present on this forum regarding the number of processors that I can use (Choosing an Appropriate Number of Processors) to run the simulation. However, the mode run time while using 40 processors was 28 hours (approx.) and while using 16 processors was 22 hours (approx.). My simulation period is of 15 days over 3 one -way nested domains.
Also, according to the rule

For your largest-sized domain:
((e_we)/100) * ((e_sn)/100) = least amount of processors you should use.
For your smallest-sized domain:
((e_we)/25) * ((e_sn)/25) = most amount of processors you should use.


For my largest domain e_we = 100 and e_sn = 104 & for the smallest domain e_we = 169 and e_sn = 187. Accordingly, I should be able to use 2 - 50 processors without any problem.


Here is the system config:
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
Address sizes: 48 bits physical, 48 bits virtual
CPU(s): 48
On-line CPU(s) list: 0-47
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 1
NUMA node(s): 1
Vendor ID: AuthenticAMD
CPU family: 25
Model: 8
Model name: AMD Ryzen Threadripper PRO 5965WX 24-Cores
Stepping: 2
Frequency boost: enabled
CPU MHz: 1800.000
CPU max MHz: 7021.0928
CPU min MHz: 1800.0000
 
More number of processors cannot always ensure faster model integration. This is because the communication between different processors will cost time.

Please follow the method Kelly recommended to determine how many processors are appropriate for your case.
 
More number of processors cannot always ensure faster model integration. This is because the communication between different processors will cost time.

Please follow the method Kelly recommended to determine how many processors are appropriate for your case.
@Ming Chen
Thank you for the clarification.

With Regards,
Tanmoy
 
Top