Large Model Performance

[anderson51]

Hello,

I am looking for hardware recommendations to increase performance on models that are in the 150-200 MB size range? Experiencing quite a bit of lag when zooming, orbiting, panning, especially when any of the visual styles are being used.

Wondering if there are any settings that can be adjusted, video card or otherwise, that will boost performance?

Thanks for the guidance,

Anderson

Windows 7 Pro – 64OS- (SP1)

Intel® Core™ i7-4770 CPU @ 3.40GHz

32GB RAM

NVIDIA Quadro K600 1GB DDR3

WDC WD10EZEX-75ZFA0 SCSI Disk Drive

[anderson51]

Bueller...........................Buuuuuueller......................................Buuuuuuuuuuuuuuuuuuueller.................................................................................................................................................................Anyone..........................................Anyone..........................................................

[jggerth]

SSD drive to start with, then overclocking and watercooling.

Anything elses is just going to be workflow tweaks -- Xref management, partial loading , layer & spatial indices, freezing vs off.kill as much eye candy as possible.

[anderson51]

I was thinking going the SSD route or adding RAM to the video card would help....

150 MB model is a relatively modest size in our industry, I will be in trouble on the larger jobs....wondering if Revit or Inventor might be the way to go on those?

Thanks for the input JGerth

[dgorsman]

Not sure what industry you are in, 150 MB being a "modest sized" model sounds a little off, even in the Civil3D world.  Like JGerth mentioned data management plays a big role: level of detail (not modeling leaves on trees for a 1:5000 drawing, using external files for data storage), compartmentalization (XREFs), and re-use (blocks, XREFs).

The choice of software is highly dependant on what is being worked on.  I wouldn't use Revit to design a check valve, nor would I use Inventor to design a warehouse.

[anderson51]

Let me refrain, the job size is modest for the Commercial AG industry in designing and laying out grain handling facilities. Other than the PEMB containing some architectural details, its alot of structural steel and concrete, no real eye candy or leaves happening.

Is Vault the way to go for data managing AutoCAD?...

What is a good resource on learning the best practices for data managing CAD files?

Thanks dgorsman

[jggerth]

by eye candy I was referring to the AutoCAD 'fluff', Xref palette, properties palette, layermanager palette, ribbon.

I interpreted the 'leaves' reference to avoiding excess detail in the dwgs.  Just because a pipe is manufactured as corrugated steel, does not mean it needs to be modeled on a site plan withe every corrugation laid in.  Maybe worth doing on a detail of a specific area, but not everywhere.

Similarly, every spoke on a wheel of a grain conveyor does not need to be modeled in 3d with textures in order for an experienced contractor to build the facility.  

Just because I can, doesn't mean i should.  And while the job size may be modest, that file size isn't.

[ampster401]

I've got little exposure to Revit, but it sounds like it would work best for what you describe.

[dgorsman]

Data management depends a lot on the software in use.  If you are designing the mechanical systems in Inventor, you would be building common parts then bringing them together into finished models.  If you are designing the structural steel and concrete components in Revit, you would be using families (like the others, not intimately familiar with the product).  If you are using AutoCAD, then you would be using blocks and XREFs.

I'm involved in process facility design - kinda-sorta similar to what you do only with fluids rather than solids.  That includes piping, structural, and electrical work..  A valve for example - the critical data common to every instance of that valve is stored elsewhere and looked up using an index code.  Only enough data to identify the valve is included with the valve object itself.  A pump includes all sorts of internal components; we aren't responsible for any of those, so we only need to model the external envelope to the point where we can be certain piping and steel isn't run through it.  We commonly have multiple "trains", or systems, which are identical copies of each other.  Instead of redesigning each from the ground up we create one then use multiple XREFs of it in the overall model.

In general data management boils down to a single concept: single source of truth.  Don't create individual instances of data, instead create pointers to a single instance of that data.  If you have a standardized auger that you use in multiple places in a model, make a block of it and insert copies of that block.

[anderson51]

Thank you for the knowledgeable reply dgorsman, I appreciate it.