Learning Forum

[ahmed.desoki]

I have a structural FE model of 4.5 million nodes. Indeed the model has a lot of details that hinder reducing number of nodes. The PC I work on has 8 cores, 32 GB RAM and NVMe SSD with plenty of free space.

When I tried to run static analysis of this model, Ansys issued the following error massage informing that RAM is larger than my RAM.

[Rob]

Moved from Feedback. nLooking at the posts the high RAM is to solve the model without paging, if you don't mind paging to disc you'll need much less: hence we have SSD hard drives for this sort of problem. But you'll still need 42-ish GB to run the model. or can confirm. nAs an aside, 200GB isn't that much on a cluster box, our older large memory nodes are 300GB and and the newer ones 500GB. n

[Aniket]

1) Why Ansys requires huge memory? I think Nastran is not as memory hungry as Ansys!! (I'm an old user of Nastran, but this was before the multi-core era.)nCan't comment on this. Following link explains recommended solver for your problem size. Please note, nodes do not indicate the number of degrees of freedom (there is generally a multiplier of 3 or 6 depending on the element type used)nhttps://ansyshelp.ansys.com/account/Secured?returnurl=/Views/Secured/corp/v201/en/ans_bas/Hlp_G_BAS3_2.htmln2) Given the number of nodes and cores, can I estimate the amount of memory needed for both in-core and out-of-core analyses using the distributed sparse matrix direct solver?nWe have standardized memory usage for direct, iterative, in core, out of core solvers at the following location:nNote that in core run requires more RAM. While out of core runs will require more Hard disc space.nn3) If my RAM size is less than the memory needed for out-of-core analysis, what can I do in this case?? Should I decrease the number of cores? change the solver type? use substructuring (super elements) for some parts with dense meshes?nYes, you can reduce the nodes first, reducing cores would help too, but not by much. Yes to rest of the questions as well.nn4) I have several PC's identical to my PC connected to 1 Gbps Ethernet. Can I bypass the memory shortage problem using Distributed Memory Parallel (DMP) (assuming the sum of memory of the networked PC's are higher than the required memory on a single PC, and my license allows this)?nWhich license do you have? is it free student version? if not which is the specific university license that you are using?n-AniketnHow to access Ansys help linksnGuidelines for Posting on Ansys Learning Forumn