Mazda MX-5

Please checkout my profile page for Workstation Specs. By the time of creating this one, I had no SSD cache and only 1 X5650 at same frequency though. Project details below. . . S I M U L A T I O N D E T A I L S G r i d s i m u l a t i o n : - 400 x 8 x 400 meters real scale (XYZ), 50cm cell size. - Water depth = 4m - max wave height = 4m - 25fps - 600 frames countdown + 1102 frames timeline with file output. - 30.118.720 Grid particles per frame (In preview you only see 6.000.000+ because proxy display is used). Storage used for grids: 1.05TB (plus Proxy) Storage used for gridDisplacemnt: 18.6GB SimTime for grids: 17.6 hours for 600+1102 frames S p l a s h s i m u l a t i o n : - 5 Splash Emitters ( 4 in one run, number 5 is the center of maelstrom and includes complete foam sim for entire maelstrom ) - Splash simulated for frame 300-600 of original grid simulation - between 4 and 6 million splash particles max ( 5 emitters, average around 3.9 million ) SimTime for 4 Splash emitters ( without Center emitter ) = 39h:04m Storage used for splash: Unknown...Sequence deleted. F o a m s i m u l a t i o n : - Foam simulated for only 107 frames to test ( 300-407 ) - 24.000.000+ Foam partiles per frame (Only foam maps saved, NO particles saved to disk) SimTime for Foam + Center splash = 24h Foam, Splash unknown (sequence deleted). Storage used for 107 Foam Maps ( 8192 x 8192 ) : 40.1GB C r e a t i n g M e s h e s : Meshes for GridSimulation: - 203 meshes created ( frame 300-502 ) - Used mesher: rkMesh (renderkit meshing inside RF, NO renderkit) - Average 16.900.000 Faces Time for meshing = appr. 4.5 minutes/frame = appr. 15.2 hours / 203 frames Storage used for meshes: 178GB (including deformed output from maya) R e n d e r i n g I n C i n e m a 4 D r12 : Native Renderer used ! NO particles rendered. Only the meshes from grid simulation + Foam maps. Could not render the particles for two reasons: 1) Deforming them in maya works fine, but then I need to output a particle sequence to be used in C4D. The RF Plugin for Maya crashes when exporting the deformed sequence. 2) Even if I wanted to use the original undeformed sequence I would not be able to render it with particle shading because I only have 24GB ram. Consider the heavy meshes and maps... Filled up ram instantly when calculating the shadow maps (selfshadowing is a MUST for shading this stuff with particles). Not a problem: To make it look great even from close up (about 2 meter distance to water suface) It would take more than 90 million SPH particles. Im getting older and theres more to be done than waiting for RealFlow to complete such a sim. Doubt it could do that with 24GB ram anyways ;-) Creating meshes for the splash did not pay off...very slow and also too heavy for rendering together with the other maps. So no need to explain why I did not add extra spray and high res mist and dust for final rendering. No chance! 1920x1080, 203 frames WITHOUT any Maps = 1h:45m 1920x1080, 90 frames WITH Foam Maps = 50 minutes Storage used for Rendering: max 6GB ram + 422mb output (PNG) C O N C E P T : T h e P r o b l e m s Y o u W i l l F a c e : To simulate a maelstrom (certainly with such a scale and detail) you cannot fill up a cube with the dimensions of the BoundingBox your maelstrom will have. Well you can, but not if you dont have god-like software and uber-god-like hardware and years to finish. You want the surface. nothing more! So simulate a flat layer of fluid and later deform the output in your 3D application to shape the maelstrom. Problem: Flat layer does not behave like a real swirl inside a volume. So you need to "bend" all the forces that move your fluid (certainly the resulting forces of rotations) to match the shape of your maelstrom. Means: You dont just take a gravity and vortex and go home, but you must model the gravitational and resulting forces of rotation (and others) in a field, so that after the deformation of this field their vectors would all point straight downwards. And that is hell nasty, certainly if you cannot bend forces at all (no way to bend deamons in RF). To create an array in a 3D app providing objects and then assign forces to it in RF is not good...the more deamons you add, the slower the sim gets. Using Object fields in RF is possible but to avoid uniform patterns in your simulation at such a scale and resolution is impossible if you use low res geometry as source for the object field. If you use high res objects the sim becomes hell slow. Solution? Develop a tool to create/deform custom force fields inside RealFlow. Concept: Simulate flat layers, deform them in maya. THere u go... Took 9 days from start to finish (including 3 versions before this one). Would love to add spray, mist and dust sims and some atmosphere but no chance.

• published: 26 Jun 2011
• views: 2181