Hello ,
I am making a 2D simulation using fluent
at first i run the simulation on my i7 laptop (4cores 8 threads) than i got a z820 workstation with 2 processores (24 cores 48 threads)and i run the exact same simulation on the workstation and i was schoked that it's only 2 times faster than my laptop.
the setting i chose for the laptop and workstation are :
1/-laptop -parallel -double precision -6 processors
2/-workstation - parallel -double precision -36 processors
How in the World that a workstation in only 2 times faster than my laptop ?!!
Do you have any idea about what may the problem be ?
PS:
both the laptop and workstation have the same ansys version 2019 R3.both the simulations setting are the exact same except for the number of precessors used for parallel solving.
Many Thanks
An informative test may be to find out how long your simulation takes using 6 procs on the workstation, which may give some comparison between laptop and workstation. Then you may wish to find the scaling on the workstation alone. Run with 1, 2, 4, 10, and 36 procs. This will establish speedup.
A reason for poor scaling may be highly uneven geometry, among others.
thank you for the reply i will test that as you said but in the mean while do you have any idea about the Ansys HPC licence ?! i am not sure about what does it mean but i think it may be a reason behind that ?
Do you have any idea about how many cores does the default version of ansys support ?!
Many Thanks Sanjay Kharche
Unfortunately, I do not use commercial packages. I am a developer, and implement my own MPI codes.
I did a lot of work on parallelizing flow codes for the US DoD and DoE. It sounds a lot better than it actually is. If you pull up Windows Task Manager and select the performance tab, you can see how busy the processors aren't. Every CFD code I've ever worked with can easily peg one processor, maybe raise a second to 50%, and sporadically use a few more. The idea of pegging 4, 6, or 8 processors at 100% is a pipe dream. The center for High Performance Computing at ORNL and the new supercomputer under construction there are staffed with smart people who work very hard to find problems that showcase the power of these machine. This is the exact opposite of figuring out how to use massively parallel machines to solve urgent problems that take too long on a single processor. It's like creating very expensive bolts to fit a wrench you already have rather than a special wrench to fit any bolt.
- Badges
- Science topic
- Similar topics
- Psychology
- Mental Processes
- Learning