Liquid in a cyclone

[Svetlin.Nikolov]

Please note that there are several types of Non-Newtonian fluids and Phoenix supports just one of them - it can make the fluid act less viscous when it's moving more slowly, see here for a longer explanation: https://docs.chaos.com/display/PHX4MAX/Liquid+Dynamics

[JuhaHo]

I think I'm now getting the flow lines better. However it seems all the simulations are running into memory issues.

[Svetlin.Nikolov]

How many voxels are there in your grid?

[JuhaHo]

Thanks,

I had to close the session, but I think there was about 20 million - maybe 21.

[Svetlin.Nikolov]

This is a super low number, which means the high memory usage is not because of the fluid. Then could it be that the interacting geometries are too many? How many tyFlow particles do you have?

[JuhaHo]

I think I had disabled tyFlow. In this animation I will simulate tyflow based on presimulated PHoenix.

[JuhaHo]

Actually, I rebooted the workstation and restarted the job. Now there isn't memory issues any more.

[JuhaHo]

I don't know what to do. 14 039 712 Voxel simulation refusing to simulate due to memory consumption. What I should try next?

[Svetlin.Nikolov]

Hey Juha,

As we already mentioned twice so far in this conversation - obviously the voxel count is not the problem as it is very low.

Please answer these questions:
- Do you have tyFlow in your scene?
- Do you have any other object in your scene that creates meshes or particles? Such as pflow, thinkingParticles, Stoke, or even any imported alembic or xmesh caches?
- How many such meshes and particles of the above do you have in your scene?

It would be easiest to get an answer if you can send the scene over so we can quickly check it and return an answer to you.

If you can't send the scene, then it's very important that you make maximum effort to explain to us about everything you have in your scene - every object, every plugin, and every connection between them.

Without this information, please understand that we are shooting in the dark and can't imagine what might be happening.

Cheers!

[JuhaHo]

Thanks,

I have only tyFlow in the scene, but it's inactive. TyCache is hidden too. So, there's no particles in the scene. There's 12 geometry objects in the scene interacting with Phoenix FD. Also no foam- or splash particles.

This is the latest message from last night due to lack of memory.

[JuhaHo]

Yet another crash; the workstation booted itself during simulation. I studied event viewer and there was lots of warnings on running out of memory. I had 14 million voxels in my simulation and 96 GB RAM in my workstation. What would be suitable amount of memory to simulate 14 million Voxel system?

Please answer this, I would really need to show some signs of success to the client.

[georgi.zhekov]

Hey man,

Could it be that you have a faulty memory? 14 million voxels as Svetlin said are not that much. Maybe you can run some memory tests to rule out that there is no hardware problem - sometimes a stick of RAM can be faulty and produce errors.

If it is not a hardware issue - could it be that the objects in your scene are very heavy and consume a lot of memory on their own?

[JuhaHo]

Thanks,

I have scheduled memory check on next shut down. However, all the Event Viewer messages are related to insufficient memory. I have 96 GB RAM and 220 GB Swap space. Both of those are utilized during simulation.

[georgi.zhekov]

Then how about my other question - how heavy is the interaction geometry, could it be that it has a lot of polygons?

How about if you leave only a single object to interact with the sim?

[JuhaHo]

Thanks,

The most complex object in the simulation has 1.6 million polygons. Then there's two with 300k polygon and the rest are more like boxes.