Many Collision Objects with Liquid Sim

[DragonOfDarlH]

Hello dear Forum,

i want to achieve a crash of a shell with liquid inside.

I have a finished rigidbody / Fracture Sim with around 1000 Objects. It should filled inside with water... The setup was easy, but the simulation itself takes 6 hours for a 1 mio Grid of 600 Frames. But I don't get the desired result and the amount of iteration sims drives me crazy. Is there a way to speed up the sim or maybe a different approach? I cant lower gridsize because the result is not really comparing to a higher sim (should look a little more detailed -> 1 Mio).

Are there any tips or is it just possible with faster cpu / more sim clients in the network?

Actually using 3dmax 2016, phoenix fd 3 demo, intel 5820k @ 4Ghz with 64 GB Ram and Win10 64 Bit on an SSD.
Performance simulation with everything 0,00 - 0,01 C/s.

Key facts about the scene:

Liquidsource -> VolumeBrush -> innersphere
animated sphere -> collision object (just normal animated sphere)
Grid: Adaptive grid, maximum 1 mio cells

include list: sphere (animated fractur parts, emittersphere, sourceobj)

Thank you in Advance!
Greetings,
Dragon

[DragonOfDarlH]

Or maybe I should post the question a little bit different. Which factors are causing a dropdown of simulation speed? I made another fracture simulation with lot of objects... nothing made in the scene...
And now its simulating with 0,6 MC per sec.

Whats causing this huge difference? Scene Scale? Units? Objects in the Scene? Number of Keyframes on objects in the scene?

To know which factors are responsible for slowdown, would help me a lot. Thank you!

[Svetlin.Nikolov]

Huh, would you be able to drop the scene to Support (at chaosgroup dot com) so we can profile it and see how to make it faster? The difference in speeds also seems like a symptom for something we might need to fix. If you have the thousands objects working fast with one setup and slow with another, we definitely gotta check if something's wrong on our end.

Thanks!

[DragonOfDarlH]

Hello a0121536,

thank you for your offer. I tried to build almost exactly the same conditions as in my original scene. Then I had an idea... I used rayfire for Fracturing and i tried the caching tool. It stores the keyframes outside of the scene. The result is 1,4 - 2,6 M C/s with 5 Mio Cells and 6 steps! So the neckbreaker was the heavy keyframing of the fracture parts. The reason why my 1st testscene was much faster is the length... i didnt simulate it through the whole length of 600 frames.
So I think the key to a fast simulation with heavy number of collision objects is to reduce or outsource the animation! Now its incredibly fast and its fun to watch how the simulation goes

Thank you anyway!

[Svetlin.Nikolov]

Ah, alright, glad you caught it!