Car cad data extreme slow down

[ali_cgi]

Hi guys, I'm wondering if anyone can advise on the handling of large cad data exports?

I have an fbx of a car which is 700mb 'full car, interior + wheels + body'
This file is incredibly slow, max becomes unresponsive/ unworkable,

If i take the body on its own the fbx is 200mb,
This works fine, I can tumble around etc quite comfortably, even render with rt, but i really need the whole car in the scene to work quickly

My system is a dual proccessor quad core xeon, 24mb ram, nvidia gtx 780 3gb graphics card,

Are there any tweaks i can make to max? I've tried switching to OpenGL legacy mode but still very slow if not worse,

Does this sound like a more powerful machine is required, if so how can i pin point which components are causing the bottleneck?

My initial thoughts are the graphics card, what improvements would a good Quadro card give?

[DaveKW]

This is something I'd like to know as well as we have similar issues, not unresponsive, but slow and tedious if the full car is visible.The files are around 700mb+ fbx, 25,000+ objects, only 4mil+ tris though. I'm running a single 6 core, 64Gb ram, Quadro 4000.

[kosso_olli]

I find that Max has serious problems with large amounts of objects. 25,000+ objects is way too much. Most cars we are getting from the suppliers have up to 7000 objects, around 4000 of these objects are useless helpers I don't need anyway. I group the parts into larger elements like headlights, wheels, brakes, grille etc. Only a few dozen helper objects afterwards, around 3-4000 car parts in total, interior+exterior. Polycount around 24million, running at about 12fps on a Geforce GTX Titan. Never used a Quadro because they are too expensive and the gaming cards are faster in RT.
For RT it is best do export the groups as proxies, so that the response time when navigating the scene is lowered.

[DaveKW]

I agree, 25,000 is a crazy amount. Around 10k of that will be geometry, the other 15k are helper objects which we have to keep as that's how the assets are tracked. We mostly work on configurators and each finish is a unique piece of geo rather than have 1 set of geo and have a shader change.

[kosso_olli]

We mostly work on configurators and each finish is a unique piece of geo rather than have 1 set of geo and have a shader change.

I can see no benefit in doing it this way. Why is that? If it is required by the client, you have no other choice, I know. But Max will always be slow with these large amounts of objects, especially helpers. Instinct has written something similar in this thread over here: http://forums.chaosgroup.com/showthr...643#post668643
Also, I am afraid that no existing software package out there could handle this.
Are xref's an option for you?

[DaveKW]

It's a client thing...each piece has it's own product code that needs to be tracked. We're working on things being more streamlined, but it's interesting seeing we're not the only ones struggling with large amounts of objects.

[kosso_olli]

Interesting. Can't think of any solution, sorry. Maybe ask Instinct, they are dealing with this every day.

[ali_cgi]

Ok so i seem to be dealing with similar data lots of helpers/null's within 'helpers/null's' within 'helpers/null's' and so on,
Now its difficult to collapse with data, and remove the helpers, as it ruins the clients object tree, so if anything changed it could become a nightmare,
Although i would be very interested in what others are doing regards to simplifying scenes if the helpers are causing the slow down,
My fbx is 700mb but saved max file 2.8gb,

[Morbid Angel]

is it possible for you to post a part of your scene for investigating?

[ali_cgi]

I can't post as its confidential data so could get into serious trouble, i will try and show a scene tree, to illustrate how the file is organised

[Morbid Angel]

Im unclear is the car moving at all, or parts of it? Whats the need for the deep structure like that?

[ali_cgi]

Nope its for a still, its how it was delivered from the client, not sure if its how the cad program exports,
What would the easiest way to simplify this be?

[Morbid Angel]

hmm well for a still there are a few things you can do - kill all the nulls in the scene, should be painless. Also you can attach my material, then export to vray proxy. That will surely make the scene fast.

Those steps can be coded and repeated each time you get an update. Unless they expect that scene back from you in the same form? then its kinda a problem

[ali_cgi]

Thanks for the reply again Dmitry,

what do your mean by 'attach my material' ?

I was thinking of killing all the nulls, but would like to keep some kind of organisation by say keeping the top null, also there are various options for bumpers, different models etc, so flattening the scene entirely would make it extremely difficult to turn off parts i don't need,
Would it be a manual process of me going through each part and flattening things down?
Or are there any quick ways? Unfortunately I don't have the luxury of knowing how to code,

Nope i don't have to return the scene just a render, so i can technically do what i like with it,
Mainly trying to keep things organised for my own reasons,

[Morbid Angel]

What I usually do on complex scenes is - a bit of work now can be a lot of time saving later. So I would clean up the scene best I could, if you need to keep parts of structure, there is no way to automate that. But I would setup a main scene where those parts can be easily assigned to layers and turned on/off that way rather then using nulls or hiding them.

Attaching by material means that you can combine parts of the same group into single mesh. Max generally likes less single meshes even if they are complex. You can have a scene with a million of cubes it will be slower then having them attached into one.

So lets say you have interior where you have seats made out of leather, well you can select everything that has leather in the interior and attach it into single mesh preserving that material. Usually you don't have as many materials as you have objects so that's the point of optimizing.