My company owns both NEI as well as NX (NX is a new purchase for us). We have all run different, but very simple, models on both programs for comparison and found a noticable difference in the results. Can anyone explain to me the difference in computational method so that we can choose which program we should use on future projects for best results?
FEA codes have numerous proprietary algorithms and element formulations that differ from vendor to vendor. Contact algorithms in particular are quite different. Other items are exactly the same, the 10-node tetra element formation is the same across most vendors. In the past, we've run the same model 10-node tet model in Nastran, all flavors, ANSYS, ABAQUS, etc. and the answers are the same out to the full precision of FEA results files.
Please post one of your models with different answers and we can look at it and perhaps provide some insight on what might cause the differences.
I am currently a user of both NEi [AUTODESK Nastran] and NX and would very much like to see what models you are analyzing that have different results. I have modeled several problems, both standard linear analysis and with contact and have found the results to be very similar, although I think I find the NEi solvers to be faster even when the models have a larger number of elements ie: NEi vs NX. I look forward to seeing your analysis and your results.
I'd have to say we experience the exact opposite of what you describe in terms of performance. We typically make fairly large models (4M-5M dof or more) and we've seen our NX NASTRAN models (remember NX is the flagship CAD/CAM/CAE from Siemens :-) ) run significantly slower with the NEi solver, if running at all. There are a lot of annoyances that have been addressed in both MSC and NX NASTRAN that linger in NEi. It is quite often that we have to "dumb down" our NX NASTRAN models to send them to our customers...
As for the difference in results, the only time we've seen anything is when NEi was using cards in a different manner than MSC/NX NASTRAN... After all, the NEi code is as much a NASTRAN code as Altair Optistruct is, so there will be some differences that do not manifest themselves between NX & MSC NASTRAN.
What NEi solver are you using? There are several and it is never wise to use the default. It is interesting that you find NX faster when I find NEi faster. It would be interesting to see the models and to review the issues found.
We do not use NEi's solver, some of our customers do, and the majority of them have assumed they were running NASTRAN all along... Our analysts rely heavily on "real" NASTRAN functionnality such as DMAP that is not available with NEi's solver, so it was never a consideration for us. And we own an armada of software licenses from a plethora of different vendors just in case you assumed we were Siemens-only.
NX NASTRAN has many possible performance tweaks that can be used to increase its performance, starting with generating a timing constant all the way to GPU computing. As you said, it is never wise to use the default
We greatly respect all of our customers and assume they know what they're doing with their software. It is very typical for them to report run times that are longer than ours, and they typically run on better hardware than we do... Since we have multiple datapoints from multiple customers in different industries, we came up with a plausible trend... I'm sure there are some instances NEi's solver is faster, we just haven't experienced one yet...
As you can imagine, there is absolutely nothing we can share about any of our models, but the usual issues old time NASTRAN users will recognize, issues that have been addressed a decade ago by both MSC and Siemens, crop up often. And it goes far beyond a mere PARAM that is not supported.
One thing I can tell you for certain is that the common denominator between NX NASTRAN and NEi's solver is a model that is unfortunately not a full fledge NX NASTRAN model.