Under Criteria and Calc Options in RAM Concept V4.1.1 there is an option to select the matrix solver as either Paradiso or Skyline. Where can I get more information as to what these solvers are doing? Also we have a pretty large model and would like to solve using a smaller mesh size, however we keep running into out of memory errors, does RAM Concept just use RAM when solving the matrix? If the model is large is there any way to make it access the hard drive?
S. Bouvia
The Pardiso solver uses some matrix consolidation methods and comes from Intel. There are several web pages on the Intel domain that explain in more detail (a Google search can get you there quick). The Pardiso solver is faster than traditional matrix solvers, but yields the same results.
One consequence of using the Pardiso solver is that we can't actively correct for nodal instabilities. What some of the Ram Programs have done in the past was to substitute a tiny non-zero stiffness for a node that was otherwise free to spin allowing the analysis to complete rather than stopping with an instability message. Using a pardiso solver with the same model might yield an instability. Not that that would matter much in RAM Concept since those models are more often than not perfectly stable.
RAM Concept does not offer an out-of-core solution where results are written to the hard drive rather than stored in memory, so for very large models you can run out of memory. Below are some tips I commonly give to help in those cases:
1. Remove any embedded drawing files (File – Delete drawing)
2. Mesh as large as possible while still getting the accuracy you need.
3. Remove non critical design strips, sections, or loadings.
4. Stop other applications.
5. Sometimes you can get a model to run once, but not twice, in those cases, close the application between runs.
6. Load History calculations (and Mat footing analysis) are iterative, in those cases adjust the convergence tolerance under criteria – calc options to reduce the cycles.
Thanks for the answer, however I have a few more questions.
1. How much memory can RAM Concept actually read, for instance I have 16 gb of RAM on my computer. At the bottom left of the screen the program says Used: 4095 MB, Available: 3736 MB. My resource monitor shows 4188 MB used and 12075 MB standby., why can RAM Concept not use the standby, note that 3MB is shown as hardware reserved as well. I mean ideally you could solve as large of a problem as the program could use RAM for, because I am assuming it is accessing the RAM on the computer.
2. Will RAM Concept ever be able to use more than one core for its analysis? Currently I have a slow quad core at just 2.0 ghz which if I could use all four cores would decrease analysis time.
Feel free to email me if you do not want to post all of the answers to these questions on the forums or if there is any reading you would think would be useful to answer these questions.
1. We are currently limited to the memory on one core, so that's probably why you are seeing total memory available less than what you actually have.
2. We have some speed improvements to make on our end, and we need to make a true 64 bit version. Both of those things will go a long way to speeding things up. Beyond that, multi-core or multi-thread processing is the next step (sometime after version 5 it looks like).
Well, almost a year gone by, and into Version 5. Any update on 64bit support in the RAM Suite of products? RAM Frame and Concept are the 2 most frequently used products on my end where that would go a long way to improving productivity.
I can say pretty much for sure that compiling Ram Concept and RAM SS especially for 64 bit processors will be in the next major versions (mid year 2013 hopefully).