Cancel
Showing results for 
Search instead for 
Did you mean: 

memory problem

Creator
Creator
I am running out of scratch memory on a rather large frequency response calculation. The error code (1012) suggests that i expend the scratch file, but i am at a loss here. There is a reference to "TAN 4893" on the UGS website, but i don't know how to find that.
Any help would be appreciated. I am not familiar with the NAstran deck, as i perform all my actions in Femap, so if you can help, please realize you're dealing with a nitwit with regards to writing in a Nastran deckSmiley Happy
10 REPLIES

Re: memory problem

Innovator
Innovator
Dear Robmooij,
You simply need more disk space and/or more RAM memory, check your *.f06 file for the proper description of the error. Also, explain what means "rathar large" freq problem, how many nodes & elements have your model?.
Best regards,
Blas.

Re: memory problem

Creator
Creator
the model is some 110.000 nodes and a few more elements. I am trying a frequency response for an area including 4 modes of interest and a frequency range of 1.5 upto 14 Hz.
I tried changing some settings, including indicating s second scratch directory and the largest possible files in there. I am afraid i got the commands wrong somehow.
reading the F06 doesn't make me wiser unfortunately.

Re: memory problem

Siemens Builder Siemens Builder
Siemens Builder
Assuming that you have enough disk space available, the size of the nastran scratch files can easily be increased via the sscr keyword.

If that does not work, please post the .f04, .f06 and .log files.

Regards,
Jim
--
Jim Bernard
Advanced Applications Engineer

Siemens PLM Software
2000 Eastman Dr., Milford, OH 45150-2712
www.siemens.com/plm

Re: memory problem

Creator
Creator
I am beginning to beleive it is something different. It is a model of a large pipe cooler (bundles of thin-walled pipes, contained in a pipe plate and hanging in a water filled box). I exposed these pipes to an assumed vibration of 10 mm/s over the lowest natural frequencies. i applied the load (=velocity) to the nodes at the location of the base plate and requested, apart from the standard output, also velocity and stress (beam elements)
With this, i get a huge model and finally, after some 6.5 hours, a crash. When i do not request V, it seems to be working fine. Can it be that NAstran cannot handle these requests?

Re: memory problem

Siemens Builder Siemens Builder
Siemens Builder
The original error message you reported is pretty clear. When Nastran tells you that the scratch file needs to be extended, it means that the default allocation for that file has been exceeded and you need to increase that allocation.

By reducing your output request, you are decreasing the amount of data written to the scratch file. If you increase the size of the scratch file, you should be able to successfully request additional output data.

Regards,
Jim
--
Jim Bernard
Advanced Applications Engineer

Siemens PLM Software
2000 Eastman Dr., Milford, OH 45150-2712
www.siemens.com/plm

Re: memory problem

Creator
Creator
Jim,

the final file size with "only" stress as additional output is less than a gigabyte and is finished (just) within an hour. I didn't look at the scratch file size as the calculation ran over night.
For the calculation with both stress and velocity as output vecotrs, I have seen a scratch file size of 120 gigabyte. Then it appears to stand idle for a few hours followed by the calculation ending with the mentioned fatal error.
Total free hard disk memory is close to 500 gb.

Re: memory problem

Siemens Builder Siemens Builder
Siemens Builder
Please provide the .f04, .f06 and .log files from both runs
--
Jim Bernard
Advanced Applications Engineer

Siemens PLM Software
2000 Eastman Dr., Milford, OH 45150-2712
www.siemens.com/plm

Re: memory problem

Creator
Creator
Sorry, it is a bit different then I initally told you. The model ran for approx,. an hour with default output (no stress, no velocity) before reaching results.
With only additional stress vector as output, it has just finished after some 5 hours of crunching. It failed after choosing both stress and velocity as output.

I guess it is indeed a problem of over-asking. Perhaps the solution is to run output for stress and velocity as seperate models.
I am amazed at the amount of extra effort it takes to calculate "something extra" while this extra is based on the same nodal displacements. Guess i'll just have to ask my chief to buy my a cray..

thanks for the effort to help me. it at least forced me to reflect on what i was doing.

Re: memory problem

Pioneer
Pioneer
for moderate to huge files, you should use the estimate.exe command to find out how many resources are needed for the given job. Use like this: in a command prompt window, type "call C:\Program Files\UGS\NXNastran\7.1\nxn7p1\em64tntl\estimate.exe C:\folder\deck.dat" except of course change the installation path of Nastran as appropriate on your machine, and the full directory and name of the Nastran deck file. Running this command will return the estimated Ram and scratch file size required. It's only an estimation based on the DOF and solution of the deck, but it's pretty good.

You should probably edit the file "C:\Program Files\UGS\NXNastran\7.1\conf\nast7.rcf" and add the line: sscr=500gb (for example, if you got the space).