5 Replies Latest reply on Oct 22, 2010 4:24 AM by David Paulson

    Flow sim. crash sensitivity to the number of cell

    Jere Fabritius

      Hi,

       

      We are using Flow sim and SW 2009 sp 5 with WIN xp 64. Resently, flow sim. has started to crash after reaching a certain limit of cell, for ex. 250 000 fuild cells. This seem to happen in every flow sim study and in every model. Some times it crashes even with rough cell meshes, cell count only a couple of thousand. We've tried everything short of complete WIN xp reinstall, nothing seems to work. There is no change if you use network drives or the drive in the computer, closing all programs does not help. I found an older post decribing a same kind of problem, but the fixes offered there did not seem very practical. So my questions are: Are any of you experiensing the same problem and is there a fix to this or is it just something that were are supposed to live with?

       

      PS. Mayby it's the old Soviet mathematics behind it all...   :-)

        • Re: Flow sim. crash sensitivity to the number of cell
          Alec Chalmers

          You didn't mention the workstation specifics on which you are running Flow-Sim, i.e. the CPU and the amount of RAM. Flow-Sim takes advantage of multiple processors and uses large amounts of RAM. I believe that your hardware is limiting the application's performance.

          • Re: Flow sim. crash sensitivity to the number of cell
            Bill McEachern

            I wouldn't go slaggin' the "old soviet math model" - that stuff is awesome in my humble opinion.

             

            I certainly don't get things to crash unavoidably. Are you getting the dreaded "Solver abnormally terminated error" or what? If so add a bunch of goals to track the global maximums and see if something is blowing up. What does your VAR or SWX support say about this?

            • Re: Flow sim. crash sensitivity to the number of cell
              Chris Michalski

              Have you checked HDD space to make sure your page files aren't being limited?  I accidentally had numerous crashes lately when the page file filled the drive because IT limits file allocation.

               

              I would say it's hardware (RAM limited) but you state that this is a recent development.  Look for something that changed such as a program receiving an update.

               

              My models rarely have less than 500,000 fluid cells so it's not a hard stop - it's all based on memory and processor.

              Case in point, current study is 493,892 fluid, 687,243 solid, 1,335,405 partial - sure it took 20 minutes to mesh and is almost 4minutes per iteration but as long as you have the memory it's fine.  I'm pretty sure with my 12GB of RAM I've run over 4million total cells before in 2009 x64 SP3.  I see the biggest problems when IE updates itself, 90% of my crashes are IE8 related.

               

              As Bill said, what error does it bring up?  Or does it simply shut down SW?  I know it sounds weird, but I've had quite a bit of success in trying to eliminate problems using the SWRx.  Everytime I get recurring crashes I tell it to re-open in SWRx and record the problem to forward to tech support.  Something is different when Rx is running because it skips over the steps that cause failures.  I've never sent an Rx request to tech support, just opening SW through the Rx bypassed the problem.

              • Re: Flow sim. crash sensitivity to the number of cell
                Jere Fabritius

                VAR suggestions in this order.

                 

                1 Remove all add-ins Done

                2 change to a another computer Can't do

                3 close firewall/virus check programs done

                4 memory check in bios done, no problems

                5 fix sw install currently installing

                6 complete reinstall SW next on the agenda

                 

                Details about my system:  Win XP 64bit version 2003, SP 2, Xeon 5450, 8 GB ram, free space on hard drive 86 GB, page file size 2-12 GB.

                 

                SWrx does not help.

                 

                Here the error message, if it helps.

                 

                 

                 

                Fixing the SW installation did not help. However, i changed the size of a porous part in my study and after that it ran perfectly up to about 200 k cells. After that same error. I will try to do a complete reinstall, but i doubt it will help.

                • Re: Flow sim. crash sensitivity to the number of cell
                  David Paulson

                  You are not alone on this issue.  I am still on SW2009 Flow and have experienced significant crashing issues whereas I rarely ever experience crashing issues when only on SW.use.   The crashes are all related to the solver.  All of SW dissapears from my screen without any dialog box explaining the failure.  But the good news is that the data is totally recovered on restart of the application.  In my case, it seems to be always at the end of the solver's computations.

                   

                  There may be an issue with pagefile allocation.  I have set mine to a very low value becase I believe my 8 GB DRAM should be sufficient to do the relatively simple models that I design.  But  I always simplify a Flow model from an actual manufacturing model to reduce unnecessary processing time.  But with very low mesh and simple computation I still get the failures.  But totally recovered unlike what you get with a SW failure.

                   

                  I suggest that you document these failures to SW through your VAR.  Flow provides amazing results but the crashing leads to credibilty conerns.  I do not think that hardware requirements to run Flow are well known.  Including minimum DRAM requirements when running with the pagefile set to 0.

                   

                  I hoipe that your thread keeps going so that we can all discern the issues with Flow.