Big Models - How big and what hardware

[Kenny Macleod]

This is a scoping question designed to look into how powerful Flexsim is.

Obviously, the bulk of modelling activity is related to focusing on smaller pertinent issues but what if the only way to do it is to build a monster....
I believe that the size of a Flexsim model is restricted only by the computer that runs it - but just how big is that?

The wiki discussing graphics cards (http://www.flexsim.com/community/forum/showthread.php?t=272) has been of great use (thanks!) but graphics cards is only part of the story.

So can anyone answer 3 short but big questions (well, 4 if you count this one)?
What's the biggest model you have built in terms of number of objects and flow items? and what kind of hardware was used to drive it.
Can Flexsim handle and make efficient use of multiple processors and graphics cards?
Is there an ideal spec for a computer that will cope with just about anything?Note: I'm fully aware of the differences that level of fidelity and resolution make in a model so am interested in examples rather than a definitive answer. Particular interest is in Fluid and bottling models but any genre is relevant.

Thanks

Kenny

[Bill Nordgren]

Flexsim has developed a new product (not compatible with the current Flexsim product for obvious reasons) that is capable of extremely large models using distributed simulation (DS). In essence, you can build a model that is distributed over several personal computers using TCIP (local or internet).

Because models are built differently in a DS environment, Flexsim DS has a different modeling paradigm, and requires you to purchase or upgrade to the DS system. Since Flexsim DS can have any number of concurrent users the pricing structure is much higher than Flexsim. If you want pricing or more information about Flexsim DS give me a call at 801-224-6914.

Bill Nordgren

[Kenny Macleod]

Thanks Bill, I didn't know that and it's very useful to know.

What about the standard Flexsim version? I know my questions are very open but surely there must be some kind of relationship between hardware and the ability to build big models; any suggestions on what hardware spec would be required?

Kenny

[AlanZhang]

I wound suggest doing some experimentation on big models. The model can be simple enough to test only the number of objects and/or flow items.

We use CT to create models which is typically bigger than most Flexsim models. One complex model we built have more than one thousand Fixobjects and TEs, it works fine with bearable speed. Note CT can handle thousands of containers. But when containers are in blocks, they will not be normal Flexsim flowitems but data structures are used to store container data. When the container is going out of blocks, the flow item is re-created using the stored data.