0

I am running a python script on two computers of different configurations: PC-1 is more powerful than PC-2. The issue is that PC-1 has a computational time which is twice that of PC-2...obviously I was expecting the opposite!

I think that the problem comes from the scipy odeint package that I am using for solving ordinary equations in my python script. Indeed, I have used cProfile to compare the two computers, and PC-1 takes twice as long as PC-2 into the odeint function.

Here are the two configurations: PC-1:

  • 2 processors Intel Xeon E5-2609 v4 (1,7GHz, 8C, 20Mo cache, 85W), 1866MHz
  • 32Go RDIMM (4x8Go), 2400MHz, DDR4
  • AMD FirePro W4100, 2Go
  • Windows 10 Pro 64bits

PC-2:

  • core i5 6200U (2.30-2.40 GHz, 2C, 3Mo cache)
  • 8GB (1x8GB) 2133 DDR4
  • Intel HD Graphics 520
  • Windows 10 Pro 64bits

could it be a problem of multicore architecture, compilation etc?

Thank you

RomB
  • 295
  • 3
  • 17
  • 1
    The graphics card are not relevant. Unless you are running out of memory on PC-2, the total amount of RAM is not relevant. The odeint code will run in a single process, so the number of CPUs and number of cores is not relevant. So the only relevant differences in the systems appear to be the CPU clock speeds and cache sizes, and CPU-2 is a faster CPU. – Warren Weckesser Mar 13 '17 at 15:06
  • Maybe [this](http://stackoverflow.com/questions/34291639/multiple-scipy-integrate-ode-instances) will help to explain and work-around the problem. – berna1111 Mar 13 '17 at 15:06
  • thank you for your answers. I have two more questions: (i) is there anyway to overclock the CPU clock speed? (ii) How can I tell odeint to use all processors/cores? – RomB Mar 15 '17 at 08:32

0 Answers0