All posts related to V2
By Polyxo
#360702
In the meantime you might watch this Clip here to understand why a subdivision of 128 or 10000 as you have tried
in previous posts is absolutely not feasable. The respective fields should not even allow such absurd input.

Can it be that you have a CAD-background and normally don't deal with Subdivision-Surfaces?
A Rhino-User in this thread even posted a test by having dialed in 0.1 for subdivision. That's utter nonsense too.
NL could already avoid by adjusting the Slider-Input-range but I think they should introduce a little subdivision-surfaces-introduction
in their help-file too. They assume basic knowledge which is not always given.
Messing up entirely with displacement is something we CAD-People run into much more likely. Users dealing with meshes all
day are much more aware of how dividing geometry affects performance.
By feynman
#360704
Sure, like all rendermen and renderwomen, I'm into CAD ;) Because of rapid prototyping research, I can tell my Doo Sabin from my Catmull Clark and sniff the genus of a manifold mesh :D

A plane - subdivided into 4.096 triangles, with each triangle subdivided further by 128 - has 67.108.864 triangles. But still, Windows' Task Manager showed only a used memory increase by 0,8GB. Hence my suspicion that when entering 128 in the Subdivision field, the object's triangles are not further subdivided by that number.

I see you're Berlin based - this interesting research unit http://www.polthier.info/projects/index.html is actually developing many theories and algorithms on which we CADsters later base our real-life work...
By Polyxo
#360705
feynman"]Sure, like all rendermen and renderwomen, I'm into CAD ;)
Well, there are very many. The M~Rplugin for Rhino afaik is most widely used.
A plane - subdivided into 4.096 triangles, with each triangle subdivided further by 128 - has 67.108.864 triangles.
I'm pretty sure that this number is incorrect.
Given that a smooth subdivision is the goal something similar to Catmull Clark will get used internally which quadruples the face-clount
with each subdivision, given that quads are the Input-Geometry. Just stupidly cutting triangles in half might
work for a carpet but not when the input is the low-poly-cage of say a head.
Here I can not imagine Maxwell to do anything else but simulating the subdivision-scheme in the source-application and then triangulating/voxeling
the very last subdivision-level. That in essence would mean that with 128 you would end up with a far higher polycount as you put forward.
If this is incorrect, I'd be glad if someone put this straight.

Holger
By Polyxo
#360707
feynman wrote:(64 x 64 the carpet plane) x (128 x 128 subdivision input field) = 67108864 no matter what calculator one uses :)

I guess you are making things too simple here.

Create a single quad in any subd-modeller.
Triangulate it.
Now hit Subdivide. It - given Catmull Clark is used - gets cut into 6 quads.
In the next subdivision-step the 6 quads are cut into 24 quad (quadrupled)
This quadrupulation (or whatever is the proper english word) now happens in every further step.

Your modelling-application does not simply cut triangles in half.
I'm certain Maxwell doesn't do this as well.
By feynman
#360709
I'm outputting triangles, not quads, from the softwares mentioned in the beginning of the thread. When I output the carpet plane with 64 subdivisions from Alias, Rhino, MATLAB, SolidWorks, Modo, etc. - I get 4069 triangles, of course. Maxwell Studio imports this OBJ and shows 4096 triangles as expected; nothing wrong here.

Now, when I chose 128 Subdivisions in Maxwell's Material Editor, I would expect each of the 4096 triangles subdivided 128 times and Maxwell crashing - as I was told in this thread. But, because Maxwell does not crash, the 4096 triangles must be subdivided by some other figure, whatever that is. I suspect the number n one can enter in the Subdivision field does not mean n subdivisions of each existing triangle but represents an arbitrary parameter.

ps: In all Maxwell documentation and knowledge base, there is no mention of special subdivision algorithms (Catmull Clark, Doo Sabin, etc.) applied by the displacement options.
By Polyxo
#360711
Rhino, Alias and probably Matlab too do not count. They don't use a Subdision-Scheme as available in Mesh-Modellers.
Modo should by all means give you another result, given that you subdivide not as a straight mesh but with Catmull-Clark or that Pixar-thingy
(subdividing by pressing Tab I believe - I'm no Modo user).
Apart from that I was talking about what happens with you 4k of triangles after being sent to the renderer.

Here is at least how Vray deals with triangles (Loop-SubD)and quads (Catmull-Clark) as input. It to me appears most evident that Maxwell does something similar when using
pretesselated subdivision. But as you correcly stated - this is nowhere documented. So one should wait until a NL-employee shares some details.
User avatar
By polynurb
#360712
i remembered we had been here before :)

wouldn't that mean there isn't anything like meshes left at rendertime?

Tom, could you please come back for a moment...


daniel
By Polyxo
#360713
polynurb wrote:i remembered we had been here before :)

wouldn't that mean there isn't anything like meshes left at rendertime?

Tom, could you please come back for a moment...


daniel
Tom talks about On the Fly here. Pretesselated wasn't available back then.
User avatar
By polynurb
#360716
true.. i still believe that the magic in the background is somewhat similar.

or in other words i assume that there is no (as with on the fly) theoretical possibility to output the pretesselated geometry into anything like a mesh file format.
User avatar
By eric nixon
#360719
It is theoretically possible to output the pretess as a mesh, so ultimately its still a mesh.

Just to clarify each increment of subdiv doubles the number of polys.
User avatar
By polynurb
#360720
thanks for the info eric

would be quite interesting to eventually have such an export option.
  • 1
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
Help with swimming pool water

I think you posted a while back that its best to u[…]

Sketchup 2026 Released

Considering how long a version for Sketchup 2025 t[…]

Greetings, One of my users with Sketchup 2025 (25[…]

Maxwell Rhino 5.2.6.8 plugin with macOS Tahoe 26

Good morning everyone, I’d like to know if t[…]