Indigo using too much memory?

[U238]

Hi, I'm new to indigo and am still exploring.

While playing around with indigo I decided to render out a car I'd modeled in blender. The export went fine with no errors. However, when I fired up indigo to render it started eating memory like there's no tomorrow. I have 4 gig of memory and indigo caused my pf usage to jump to 8+ gig before windows finally said 'enough' and killed the indigo process. My car /is/ a little heavy (between 2-3mil polys) but not unbelievably huge and it's in a studio scene so there's really no other geometry. I've seen quite a few renders around this forum that probably are near what my polycount is so I wouldn't think that's it.

Shader-wise I've got a carpaint shader, a plastic shader, and a glass shader, and a diffuse 'clay' shader. (no textures) I'm using just a single 'sun' light.

So what could be causing this? Any help would be highly appreciated.

Thanks

[Deus]

If ( render = car ) then multiply memory = memory consumption * 10

[U238]

Sorry, I don't quite understand. Care to elaborate?

[CTZn]

I'm no blender user but have you applied modifiers (Ctrl+a i think) before exporting ?

[palawat]

Rendering HUGE image size and Supersamples can drain RAM out.

Try reducing image size (instead of 4000x3000 try 1024x768) and set Supersamples to 1, then see if the RAM usage will be reduced.

Hope this help.

[Zom-B]

how big are your exported mesh files... 3 million polys are peanuts for Inigo, so I think CTZn maybe pointing in the right direction!

also render your work via console to save a lot of RAM if your output res is (very) high

Code: Select all

cd indigo_v1.1.5
START /B /BELOWNORMAL indigo_console.exe "M:\- INDIGO TESTS -\test.igs"

[U238]

My output res is just 800x600 (for now).

I'll try applying all the modifiers then exporting. The .igs is actually quite large. (350mb ) Is that normal for 2-3 million polys + shaders? (and only one 800x600 texture)

[Borgleader]

If the modifiers arent applied, does it also affect render times? Like...if i left stuff like subsurf u-applied (so that i could play with the settings) and such. would it make the render slower or just use more memory?

[carbon]

forget about applying/leaving modifiers will save any memory.. they are applied by exporter with exactly same settings you have set, if you are using modifiers, i suppose you need them, so there is just a few things you can do:
1 - simplify your meshes where it possible without changing overall shape
2 - maybe adjust subsurf to one step lower value
3 - change 'super_sample_factor' from default 2 to 1
4 - get more ram & 64bit build of indigo

edit:
1a - use instances (alt-D in Blender) when possible (one mesh used by more objects.. eg: wheels)

[Borgleader]

64 bit build of indigo would be useless on a 32Bit system wouldnt it? (and I have 4Gb of ram ... i dont mind much about memory usage)

[Jambert]

ZomB
Peanuts for indigo, not for memory, it's not the same

Who can say today that he render a scene with 2 000 000 polys with 3Gb of ram, I'm very interesting by screen shots...

Wath is your graphic card? how can you work with 2 million polys for one car? don't undertand how you model it, subsurf is at 5??? why?

I'm using indigo for some time now and by experience, 2 000 000 polys are too much for 3 or 4 Gb of ram...

I've 3Gb and I can't render more than 700 000 polys , before indigo begin the rendering, it loads the mesh coordonates, and it uses more memory than the render himself, when it realy starts rendering (for a 800x600, maybe not for 3000x4000)

Forget your render, or try with the Ranch, or buy more ram

[CTZn]

make up your mind:

Code: Select all

SSE present.
SSE2 present.
SSE3 not found.
Using base Indigo directory path 'C:\Progs\3d\indigo\indigo_v1.1.5'.
Scene file path: 'C:\work\3d\Indigo\camera.igs'
Using working directory path 'C:\work\3d\Indigo'.
Loading Scene 'C:\work\3d\Indigo\camera.igs'...
Using linear tone mapping with scale=0.000200
extra_atmospheric: 0
sun_theta_rad: 0.93945
sun_solid_angle: 0.00006
Sky zenith luminance: 3858.81956 cdm^-2
Sun spectral radiance at 500nm: 18102053328213.33200 Wsr^-1m^-3
Loaded texture successfully.
Loading external mesh '|pPlane1|pPlaneShape1' from path 'C:\work\3d\Indigo\objs\1\-pPlane1-pPlaneShape1.obj'...
	Found reference to material 'initialShadingGroup'.
	OBJ parse took 0.08806s
	Done.
Merging vertices for mesh '|pPlane1|pPlaneShape1'...
	Initial num vertices: 81
	New num vertices: 81
	Done.
Computing shading normals for mesh '|pPlane1|pPlaneShape1'.
Subdividing and displacing mesh '|pPlane1|pPlaneShape1', (num subdivisions = 16) ...
	Doing subdivision level 0...
		num triangles subdivided: 65
		num triangles unchanged: 63
		resulting num vertices: 191
		resulting num triangles: 358
		Done.
	Doing subdivision level 1...
		num triangles subdivided: 213
		num triangles unchanged: 110
		resulting num vertices: 534
		resulting num triangles: 1000
		Done.
	Doing subdivision level 2...
		num triangles subdivided: 783
		num triangles unchanged: 179
		resulting num vertices: 1756
		resulting num triangles: 3352
		Done.
	Doing subdivision level 3...
		num triangles subdivided: 2940
		num triangles unchanged: 371
		resulting num vertices: 6261
		resulting num triangles: 12174
		Done.
	Doing subdivision level 4...
		num triangles subdivided: 11435
		num triangles unchanged: 696
		resulting num vertices: 23615
		resulting num triangles: 46479
		Done.
	Doing subdivision level 5...
		num triangles subdivided: 44982
		num triangles unchanged: 1454
		resulting num vertices: 91530
		resulting num triangles: 181425
		Done.
	Doing subdivision level 6...
		num triangles subdivided: 178172
		num triangles unchanged: 3210
		resulting num vertices: 359803
		resulting num triangles: 715941
		Done.
	Doing subdivision level 7...
		num triangles subdivided: 448277
		num triangles unchanged: 267621
		resulting num vertices: 1076408
		resulting num triangles: 2060772
		Done.
	Doing subdivision level 8...
		num triangles subdivided: 520441
		num triangles unchanged: 1540288
		resulting num vertices: 2001041
		resulting num triangles: 3622095
		Done.
	Doing subdivision level 9...
		num triangles subdivided: 209726
		num triangles unchanged: 3412326
		resulting num vertices: 2413250
		resulting num triangles: 4251273
		Done.
	Doing subdivision level 10...
		num triangles subdivided: 29508
		num triangles unchanged: 4221722
		resulting num vertices: 2470092
		resulting num triangles: 4339797
		Done.
	Doing subdivision level 11...
		num triangles subdivided: 6168
		num triangles unchanged: 4333586
		resulting num vertices: 2482632
		resulting num triangles: 4358301
		Done.
	Doing subdivision level 12...
		num triangles subdivided: 11171
		num triangles unchanged: 4347087
		resulting num vertices: 2505230
		resulting num triangles: 4391814
		Done.
	Doing subdivision level 13...
		num triangles subdivided: 21890
		num triangles unchanged: 4369881
		resulting num vertices: 2549356
		resulting num triangles: 4457484
		Done.
	Doing subdivision level 14...
		num triangles subdivided: 43264
		num triangles unchanged: 4414177
		resulting num vertices: 2636503
		resulting num triangles: 4587276
		Done.
	Doing subdivision level 15...
		num triangles subdivided: 85974
		num triangles unchanged: 4501259
		resulting num vertices: 2809470
		resulting num triangles: 4845198
		Done.
	Done.
Building Mesh '|pPlane1|pPlaneShape1'...
	2809470 vertices (64.304 MB)
	4845155 triangles (129.380 MB)
	BIHTree::build()
	calcing root AABB.
	done.
	Root AABB min: (-2.500000,0.000980,-2.500000)
	Root AABB max: (2.500000,0.140980,2.500000)
	Allocing and copying intersect triangles...
	intersect_tris mem usage: 221.794 MB
	done.
	Build Stats:
		total leafgeom size: 4845155 (18.483 MB)
		total nodes used: 6929581 (105.737 MB)
		numInternalNodes: 3464790
		numLeafNodes: 3464791
		num_under_thresh_leafs: 1391446
		num_maxdepth_leafs: 2073345
		num_inseparable_tri_leafs: 0
		meanNumTrisPerLeaf: 1.39840
		maxNumTrisPerLeaf: 3
	Finished building tree. (Build time: 13.98289s)
Done Building Mesh '|pPlane1|pPlaneShape1'. (Time taken: 13.98603 s)
Building Object Tree...
	2 objects.
	calcing root AABB.
	AABB: (-2.500000, -2.500000, 0.000000), (2.500000, 2.500000, 0.140980)
	max tree depth: 4
	reserving N nodes: 2(16 B)
	total nodes used: 1 (8 B)
	total leafgeom size: 2 (8 B)
Finished building tree.
AutoFocus: setting camera focus distance to 1.66268 m.
Master buffer size: 12.456 MB
Settings:
	Frame upload period: 180.00000 s
	Image save period: 120.00000 s
	Splat filter: mn_cubic, blur=0.60000, ring=0.20000
	Downsize filter: mn_cubic, blur=0.60000, ring=0.20000
	Render region: false
	Supersample factor: 1
	Metropolis: true
	Bidirectional: false
	Hybrid: false
	Ray origin nudge distance: 0.00010 m
	Aperture diffraction: true
	Post-process diffraction: false
Auto setting number of threads to 2.
Finished initialisation
Starting threads... (num threads: 2)
Starting render thread. (RNG seed=2)
Starting render thread. (RNG seed=1)
Doing initial warmupDoing initial warmup.................................

2 809 470 vertices (64.304 MB)
4 845 155 triangles (129.380 MB)
1600x1200
Memory footprint: 637Mo

Scene, texture and information in WIP > Displacement tests thread.

[Borgleader]

http://blenderartists.org/forum/showthr ... t=stargate

Grab this model from LOTR Junkie, im garantying you indigo will go KABOOM.

Gotta take out all the lights and controls though or else you get errors

[Jambert]

CTZn

Thx CTZn, so using indigo displacement use less ram than using blender displacement? I'm right??

Does simple sudbivision smoothing replace subsurf?

Sorry I don't realy understand how does it work, I learn blender alone so I don't understand all news..

Thx in advance

[CTZn]

Yes Indigo uses less ram if optimization is enabled for displacement. In my image above it would not subdivide a triangle further if it's under 3 pixels in width, even if the initial subdivision setting would have required the triangle to be subdivided. In that fashion, by limiting tris subdivision, Indigo saves memory.

I'm no blender user so can't tell more

now for more indigo-oriented users:

in the code reported above, you can see that the 4 millions tris are reached at level 9, then to level 16 much less tris are added, but they are thinner at each iteration, so the gain feature-wise is noticeable, at a low expense (593 925 new triangles from level 9 to 16). That's the benefit from pixel width threshold optimization.