Skip to main content
B
bmüller1-Visitor
1-Visitor
November 11, 2014
Solved

hardware optimization - Creo Simulate 2

  • November 11, 2014
  • 1 reply
  • 3888 views

HI!

colleagues started to use Creo Simulate 2.0.

now the they complain it's too slow.

I'm the IT guy and have not much simulation experience. Only know that simulation lasts long 🙂

Hardware:

HP Z230 with

  • Quad Core i7 @3.4Ghz
  • 16GB
  • Quadro K2000
  • SSD

It's a 15x15cm EPS part.

Simulation lasts about 1,5 hours

PTC Help:

if the rate of calculation time to cpu time is near 1 (or <4), there will be no benefit from more cpu cores.

.rpt

calculation time: 5450

cpu time: 5030

--> rate ~1,08...

.sst

calculation and cpu time rates are all near 1

.pas

recommended min. solram 50 (value unit is megabyte??)

Distributed computing:

I could not find much info about distributed batch processing. As far as I have read until now, there is a built in tool to distribute the simulation over multiple workstations. How can I use this? Is there any benefit on small parts? Default 1Gbit Network.

in my opinion more cores or RAM will have no significant speedup. Distributed computing would decrease the performancve for this small sample.

  • Hardware Option: more CPU Speed (overclocking...)
  • Software Option: model optimization

is my assumption reasonable?

br Bernhard


This thread is inactive and closed by the PTC Community Management Team. If you would like to provide a reply and re-open this thread, please notify the moderator and reference the thread. You may also use "Start a topic" button to ask a new question. Please be sure to include what version of the PTC product you are using so another community member knowledgeable about your version may be able to assist.
Best answer by JonathanHodgson

Hallo Bernhard,

There have been some good discussions about this here, but I'll try to write a short answer to summarise.

Firstly, exactly what CPU is it (e.g. i7-3770)?

Secondly, the SOLRAM value can be much higher. Basically, the guideline of using half your total machine RAM is still about right, so on a 16 GB machine try 8192 (yes, it's in MB) - although the 'knee point' for solve time is probably below this.

Thirdly, where are the temporary and results files being written - are they on the SSD? If not, they should be!

Finally, multi-core usage depends on the type of analysis. If it's linear static then Mechanica (Simulate) uses multiple cores quite well, although not for 100% of the analysis. It's common for me to see CPU time much greater than the elapsed time. However, some other analysis types such as Large Displacement only seem to use one core.

1.5 hours seems like a long solve time to me, but it very much depends on the type of analysis and the complexity of the mesh. However, the fact that CPU time is noticeably less than elapsed time suggests there may be some bottlenecks.

Can you share any of:

  • the .rpt file (and the .pas and .stt)
  • a screenshot of the model, ideally showing the mesh
  • the model itself, for others to test on their own machines?

1 reply

J
JonathanHodgson13-AquamarineAnswer
13-Aquamarine
November 11, 2014

Hallo Bernhard,

There have been some good discussions about this here, but I'll try to write a short answer to summarise.

Firstly, exactly what CPU is it (e.g. i7-3770)?

Secondly, the SOLRAM value can be much higher. Basically, the guideline of using half your total machine RAM is still about right, so on a 16 GB machine try 8192 (yes, it's in MB) - although the 'knee point' for solve time is probably below this.

Thirdly, where are the temporary and results files being written - are they on the SSD? If not, they should be!

Finally, multi-core usage depends on the type of analysis. If it's linear static then Mechanica (Simulate) uses multiple cores quite well, although not for 100% of the analysis. It's common for me to see CPU time much greater than the elapsed time. However, some other analysis types such as Large Displacement only seem to use one core.

1.5 hours seems like a long solve time to me, but it very much depends on the type of analysis and the complexity of the mesh. However, the fact that CPU time is noticeably less than elapsed time suggests there may be some bottlenecks.

Can you share any of:

  • the .rpt file (and the .pas and .stt)
  • a screenshot of the model, ideally showing the mesh
  • the model itself, for others to test on their own machines?
    B
    bmüller1-VisitorAuthor
    1-Visitor
    November 11, 2014

    Hi Jonathan,

    solram gave a good boost (->40-50Minutes). We will test a little bit more next monday and update the thread.

    I have skipped the very interesting middle part of this articel... (I thought it's too old.)

    http://communities.ptc.com/thread/33467

    br Bernhard

    Terms & ConditionsCookie settingsAccessibility statement