Model speed not increasing with additional CPU cores

D

danski

New member
I recently upgraded from a Ryzen 2700x (8 cores) to a Ryzen 5950x (16 cores). I am not seeing anywhere near the performance increase I was expecting. In fact, the model runtime does not improve beyond 8 processes on the 5950x. Given how WRF is able to scale well beyond this I find this unusual. Does anybody have any suggetions on identifying bottlenecks or other issues?

  • WRF 4.4.1 (dm) compiled with gcc/gfortran
  • Ubuntu 22.04.1 LTS
  • 32gb 3200mhz ram
  • 1 TB ssd
  • SMT off.
 
Did you change the mpirun -np (value) when you increased your cores?

For example

mpirun -np 6 ./real.exe -> mpirun -np 14 ./real.exe
 
Yes I did. I wonder if this has anything to do with using Ryzen consumer hardware. Does anybody have experience using Ryzen processors?
 
Last edited:
danski said:
Yes I did. I wonder if this has anything to do with using Ryzen consumer hardware. Does anybody have experience using Ryzen processors?
Click to expand...
Can you attached your namelist.input and name list.wps so I can look at it.
 
@danski, Unfortunately our group does not have any experience using that hardware, but it's possible someone else who has will see this post and have some insight.
 
You must log in or register to reply here.
Back
Top