CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   CFX (https://www.cfd-online.com/Forums/cfx/)
-   -   Parallel Trouble: CFX 11 XP64 - Help? (https://www.cfd-online.com/Forums/cfx/70329-parallel-trouble-cfx-11-xp64-help.html)

bbmorales November 20, 2009 10:53

Parallel Trouble: CFX 11 XP64 - Help?
 
Hi everyone.

I´ve been trying to run CFX 11 in this configuration:

- Intel i5 570
- Windows XP 64bits

In the beggining, the solver just gave the "no partitioner found" or something like this.
Then, I replaced the x86info.exe for a newer (like, 1 year ago) file I had. Then, this message was gone.
Althought, the solver, when it start to run parallel - both MPICH2 and PVM - it give me a 255 error, no matter what I do.
Serial runs fine.

Does anyone had something like it? Any help?

Thanks in advance.

ghorrocks November 21, 2009 05:10

This is a known bug in V11 where it cannot recognise recent CPUs and therefore does not know what solver to use and bombs out with the error you report. There is a fix available (talk to your support person to get it) which allows it to run on more recent CPUs but the i5 chips are very recent and I suspect the fix has not been applied to these chips. Time to upgrade to V12 I think, this will work on the latest CPUs.

bbmorales December 4, 2009 10:34

Hi, thanks for your help.

I´ve updated to v12, Win XP64, but I´m having some issues.

1) MPICH2 doesn´t works. MPICH.
2) Sometimes, seems that the processor mishandles the processes priorities and then, if i´m running a 4 partition problem, CFX "forgets" to run some of them. Then, the CPU usage is like 100%, 0%, 100%, 0% and the run doesn´t work.

Do you know any solution or tip for these problems?

Thanks

ghorrocks December 5, 2009 04:59

Make sure you are not using hyperthreading. CFX does not work well on hyperthreading.

The solver should not "forget" to run processes. If it is not hyperthreading stuffing things up it could be that you have run out of memory, partitioning problems or you could be running a fancy-pants physics model which does not parallelise well (eg monte carlo radiation)


All times are GMT -4. The time now is 08:59.