Popen in multiprocessing.Pool is extremely slow

Question

I'm trying to run multiple ngspice simulation in parallel. So I have a function _runNgspice that executes ngspice.

import pandas as pd
from subprocess import PIPE, Popen
from multiprocessing import Pool

def _runNgspice(param):
    runDir=param['runDir']
    spiceFile=param['spiceFile']
    logfile= os.path.join(runDir, 'run.log')
    with open(logfile, 'wb') as nglog:
       ngcommand=['ngspice', '-b', f'{spiceFile}']
       with Popen(ngcommand, shell=False, cwd=runDir, stdout=PIPE) as proc:
            nglog.write(proc.stdout.read())
    result=pd.read_csv(os.path.join(runDir, 'result.csv'))
    return result        

To run multiple instances of _runNgspice in parallel, I do the following:

p=Pool(processes=processes)
with p as pool:
    result=pool.map(_runNgspice, param)

This code works very well on my laptop (CPU: i7-1165G7, Linux kernel 5.15, python 3.9.7). However on my desktop system (CPU: i7-12700, Linux kernel 5.17, python 3.9.7) this is extremely slow. What I find interesting: if I run _runNgspice without pool.map all is fine. As soon as it is executed via pool.map it gets extremely slow. Also interesting: If I execute time python3 investigate.py I get a result like this:

real    0m21,267s
user    0m1,271s
sys     0m2,931s

There is almost no user time. Any ideas what could be the cause for the issue I'm observing?