Stuck? 
support@mecway.com
support@mecway.com
Mecway
FEA
Mecway/CCX solver questions
Hi,
I am new to Mecway and very impress by the capabilities with respect to price. I am trying to see what is the largest model size I can run on my harware and get a feeling of model size vs compute time. I have run into a few issues and would like to know if there are workarounds, different Windows configuration to be tried or if these are software limitations. I am running a core i7-4820 with 32 GB of RAM under win8.1 64 bit.
1- With TET10 elements, I have system out of memory errors when the solver requires more than 16 GB of RAM coming from Mecway and/or Windows. There is still plenty of free memory to allocate and disk cache files never appear in the temp folder. The solver can sometimes recover and finish solving the model, but this nonetheless requires a reboot as it messes up the display driver.
2- With HEX20, the solver is able to allocate more than 16 GB of RAM and use disk caching, but does so even when there are still GBs of free memory available. The performance drop is likely small, but can it be set to only use caching when out of RAM?
3- Mecway only uses up to 4 threads on my system, can this be adjusted or is this the maximum?
4- Is there a way to see "bad elements" other than the red x that appears on the elements, like through a list? I had error messages from CCX about negative Jacobian elements (created with the tet automesher) that were below the surface and not showing up in Mecway. I was able to spot them with CGX bad element display.
5- Where should I put the OMP_NUM_THREADS variable in the CCX input file to enable multithreaded solving? Within Mecway, it does either nothing or crash CCX.
Lastly, would any of these issue be helped with either upgrading to Win10 or the version 5 beta?
Regards,
Simon
I am new to Mecway and very impress by the capabilities with respect to price. I am trying to see what is the largest model size I can run on my harware and get a feeling of model size vs compute time. I have run into a few issues and would like to know if there are workarounds, different Windows configuration to be tried or if these are software limitations. I am running a core i7-4820 with 32 GB of RAM under win8.1 64 bit.
1- With TET10 elements, I have system out of memory errors when the solver requires more than 16 GB of RAM coming from Mecway and/or Windows. There is still plenty of free memory to allocate and disk cache files never appear in the temp folder. The solver can sometimes recover and finish solving the model, but this nonetheless requires a reboot as it messes up the display driver.
2- With HEX20, the solver is able to allocate more than 16 GB of RAM and use disk caching, but does so even when there are still GBs of free memory available. The performance drop is likely small, but can it be set to only use caching when out of RAM?
3- Mecway only uses up to 4 threads on my system, can this be adjusted or is this the maximum?
4- Is there a way to see "bad elements" other than the red x that appears on the elements, like through a list? I had error messages from CCX about negative Jacobian elements (created with the tet automesher) that were below the surface and not showing up in Mecway. I was able to spot them with CGX bad element display.
5- Where should I put the OMP_NUM_THREADS variable in the CCX input file to enable multithreaded solving? Within Mecway, it does either nothing or crash CCX.
Lastly, would any of these issue be helped with either upgrading to Win10 or the version 5 beta?
Regards,
Simon
Howdy, Stranger!
It looks like you're new here. If you want to get involved, click one of these buttons!
Comments
Can you show us the models that needs such a big quantity of elements? By the moment my main problem is not due to the quantity of elements, but the quantity of iterations on nonlineal analysis. I'm with a temporal replacement computer, a very old office notebook with i5 and 4GB ram (video card... what is this?) with a vinyl disk, so dealing with results of more than 400-500Mb in size can be slow to moving between iterations. I will try with decimation to save less iterations (but with the fear of loosing some interesting details such as the suddenly bukling or start of contact).
Regards
i started a thread awhile back wrt to model size limits. you can check it out here:
http://mecway.com/forum/discussion/99/how-many-nodes-can-you-work-with#latest
it would be great if you added your results to that thread.
I'll put a table together when I have more results, but right now I would say that everything below300 000 nodes with parabolic elements is fine for me.