6 Replies Latest reply on May 30, 2016 1:05 PM by Felice Pasinetti

    HI PERFORMANCE WORKSTATION

    Ajay S.

      Dear All,

      Good Day,

       

      I am using SW Simulation extensively for Large Assembly FE Analysis (Static, Fatigue, Optimization, Dynamics, Non Linear) and looking for Hi Performance Workstations and wondering if you can suggest  [DELL] Technical specifications to account for,

      1. Solvability [10+ Million DOF]

      2. Parallel Processing [3 projects simultaneously]

      3. Run Time [with 32 gb Ram, currently it takes 8 hours for linear static with shell element]

      4. Up Time [To overcome Hanging/Crashing of s/w]

          (enclosed sample spec. for reference)

       

      Thanks-in-advance,

      Ajay

        • Re: HI PERFORMANCE WORKSTATION
          Barry Shapiro

          Ajay,

          I would go with a 2 CPU set up with at least 12 cores in each and over 3GHz if you can afford it. Simulation and rendering both use every core you can give them and there is a 1-1 ratio to cores and time. I.E. 2 is twice as fast as 1. Therefore 2-12 core or higher CPU system would give the best result. As you go higher in core count the cost goes up and the processors slow down (GHz), so don't go too high of a core count. That's why I suggest a dual processor system.

           

          Also I highly suggest you go with at least a SSD drive for your boot drive. Everything else you picked looked good.

           

          Barry

          • Re: HI PERFORMANCE WORKSTATION
            Felice Pasinetti

            Hi Ajay,

            about core number, I suggest you this very interesting reading https://www.pugetsystems.com/labs/articles/Solidworks-2016-Multi-Core-Performance-741/

             

            In structural analysis it seems that a second processor will slow down the solver...

            "in most cases having a high frequency 4-6 core CPU is going to be much better than a higher core count, lower frequency CPU. In fact, we actually saw a steady drop in performance as we utilized the second CPU. In other words, for FEA simulation you can get some benefit from having more cores, but you should strongly avoid dual Xeon systems"

             

            I strongly agree with the SSD suggestion.

             

            Regards

            Felice

             

              • Re: HI PERFORMANCE WORKSTATION
                Barry Shapiro

                Felice,

                I understand what your saying but I have also read contrary testing on dual CPU systems.

                You can read here......I'm sure there is more if you search.

                single i7 MUCH faster than dual xeon E5-2650 v3 !!! -- CFD Online Discussion Forums

                 

                Like I said, As core count increases, processor speed decreases. So you need to calculate the best balance. The link I pointed shows that you can have bottle necks if your complete configuration is wacky. For this configuration, less RAM models is not a good thing. I see that the test from Puget systems used 1 16 gig RAM module which didn't give the second proper access to ram. This bottleneck slowed the system down.

                 

                In theory the 2 CPU system should be "almost" 2 times as fast for multi threaded applications. This is completely true for rendering and "should be" true with simulation.

                 

                Please note that I said "should be". Individual results may vary. ;-)

                 

                I know for our workstations where we are just using Solidworks and no rendering or FEA, we switched to a high clock speed i7 because Solidworks is only utilizing a single core so the i7 excels at that.

                  • Re: HI PERFORMANCE WORKSTATION
                    Felice Pasinetti

                    hi Barry,

                    thank you very much for the interesting link! I didn't know anything about the relation between RAM & CPU! It will be usefull for my next workstation!

                     

                    Anyway, I believe that, in principle, you are right, but results are strongly affected by how much the specific solver is optimized for parallel computation.

                    In your link they are talking mainly about FLUENT (CFD solver), wich I know is strongly optimized for cluster computing.

                    I don't know if SolidWorks FEA solver is optimized so much (if I remember well a few years ago it couldn't even use multiple core).

                     

                    Please let me also remark that in Puget test they used 8x 16GB RAM... so no bottleneck should occur here.

                     

                    I've searched for other test specific for SolidWorks FEA, but at the moment the Puget one seems the most reliable to me.

                    I'll keep searching as you put some doubt in my mind.

                     

                    Regards

                    Felice

                • Re: HI PERFORMANCE WORKSTATION
                  Ajay S.

                  Guys...

                  Can you throw in some light on my attached configuration please.... Any modifications required ? like,

                  1. Relationship between Ram ports--Vs--Cores---Vs--Processor ?

                  2. FEA only will b used in this Hardware and NOT CFD, hence how spec might differ ?

                  3. Graphics card should be of same type or two different type ? (for 8 GB).

                   

                  Many Thanks !!!