schwungsau

the whole pint of physical correct shading system is, you don't have hack/write shaders anymore -->setup real world attribudes and done --> artist friendly..... of course is less fun then so digg into shading models and writing own stuff....

slicing through geometrie and mirror and merge them on the edges:

testing animation parameters.... looks like growing (which are fractals too)... some issues with position corrections and need work out correct Velocity / motionblur

going deep....

more chaotic fractals..... tuned in the random values stronger this will be last one of this series.... i am going to the next level and meshing more into Z axises...

mantra spline rendering, really fast rendering....

the renderers have get updates for the 1080... this could take a while and how good it is, we will see.... GTX 980ti are cheap and works well renderers....

I7 give better performance in application because they running with 4GHz etc.... multicore Xeon 2-3 GHz are 20-40% slower with application slower but with a lots cores, Xeon rocking for heavy multicore task like rendering etc...... btw. i my expericence boosting sims with CL has low impact.

older XEON ehh? great idea..... this changes the game a bit.... i may over thing my next workstation investment....

xeon for 3000 euro workstation ?! forget it.... I7 quad or octa core is 4 times faster then xeon with some cores. xeon are expensive but they can run 24hours/7gays for years with getting warm. if u want the extreme CPU render workstation dual Xeon with 44 cores is the fastet (7000 and more just for processors) if i wanna render in octane or redshift, get a decent i7 and GTX1080.....

i did test in mantra, keyshot and modo. they all can handle to render billions polygon'e as instances. mantra take a long time in pre-render-processing but it handles it well they downside it mantra itself. rendering physical based takes a long time. Modo alos takes long pre-render-passes before it renders, but modo uses point clouds (like vray) with renders fastest, but less physical (which is still good).keyshot handles it best, fastest realtime preview and physical correct but rendertime are slower then modo/vray, but still much much faster then mantra. i need get to get a new machine with a decent graphics card and try redshift or octane. i am testing to render 3d fractals as points and splines directly in mantra. much faster rendering and less pre-render-processing but also different look and no physical renderings or GI stuff etc... mantra spline rendering:

file size ? depends.... couple of megabytes... thats all render instances --> they don't take any diskspace or memory. thats why i start with a simple meshed shape and copy / translate it hundered/tousend times across a 3d space. the renderer generate the polygons on the fly. (this are 10 billions rendered polygon and actually polygon in viewport) sometimes, i mesh out complex model with serval millions polygon's and scatter instances around the scene. in this case the instance geometry take couple GBytes of diskspace. i doing the reversed way, start simple and going extreme with instances. if you use the mandlebulb method, you get hundred billions of polygons and you have to extract /simplifyy data to be able handle it on normal computers. that why software like mandlebulber does not generate 3d data, they calculate pixels which looks like a 3d render. with math tricks it calculates normal and zbuffer, but no polygon's or camera data etc... in the mandlbulb movies the camera zooms endless, from km units to meter units to nano-meter units and so on.... impossible to handle this scale in opengl 3d package. I am planning to simulate zooms with scaling instance geo, based on keyframe and camera movements.

simple... polyline --> add noise, cpoied it multiple times with offset in pivots and rotation. copy the result multiple times, with random offset in transation, from result creatre a mesh with wireframe node, create primitive ot of it, copy the primitve multiple times with some random offset, repeat, repeat repeat..... the rendering was done in keyshot. fastest cpu tracer s far....

10 billions poly's