Large Scale Pyro Sim Cache Fills Ram Problem

[pizola 0]

32GB Ram 3200mhz

ryzen 5 1600 

Hi again guys.

Basically I have been struggling a lot with cacheing a simulation. I´ve tried .bgeo .sim and multiple resolutions on the pyro

I would hate to keep lowering the res because of ram issues. I diseabled "Cache Simulation" and I still keep on feeling my Cache. 

I dont mind waiting days for the cache if necesary (it is a big simulation, probably 40 meters long I guess) and to be honest, I might

not know in depth how to control pyros. Hopefully you can help me here

[pizola 0]

Just now, pizola said:

32GB Ram 3200mhz

ryzen 5 1600 

Hi again guys.

Basically I have been struggling a lot with cacheing a simulation. I´ve tried .bgeo .sim and multiple resolutions on the pyro

I would hate to keep lowering the res because of ram issues. I diseabled "Cache Simulation" and I still keep on feeling my Cache. 

I dont mind waiting days for the cache if necesary (it is a big simulation, probably 40 meters long I guess) and to be honest, I might

not know in depth how to control pyros. Hopefully you can help me here

 

Nave.hip6.hip

[jimeng20 7]

Hmm.. I don't think waiting days of siming is a correct thing to do. Because in the company you don't have days for only one version finished. And no matter how big your scale is, 100 meters even 1000 meters, It's not the division size defining the resolution, it's your voxel count. So I would say the voxel count above (400, 400, 400) is a relatively high res sim, which can be a division size of 10 if you are siming hundreds meters of pyro/smoke/cloud. I would suggest try to monitor the voxel count in your pyro output to determine if your sim is high res or low res.

I can't see your alembic file, so I think you could first find the mid res range first (something like(200, 200, 200)) and to get to the high res and to see if 0.2 division is way overkill which lead to the ram swapped.

Hope this could help.

 

Jie

[pizola 0]

16 hours ago, jimeng20 said:

Hmm.. I don't think waiting days of siming is a correct thing to do. Because in the company you don't have days for only one version finished. And no matter how big your scale is, 100 meters even 1000 meters, It's not the division size defining the resolution, it's your voxel count. So I would say the voxel count above (400, 400, 400) is a relatively high res sim, which can be a division size of 10 if you are siming hundreds meters of pyro/smoke/cloud. I would suggest try to monitor the voxel count in your pyro output to determine if your sim is high res or low res.

I can't see your alembic file, so I think you could first find the mid res range first (something like(200, 200, 200)) and to get to the high res and to see if 0.2 division is way overkill which lead to the ram swapped.

Hope this could help.

 

Jie

Thanks a lot. I ended up waiting 25 hours and controlling the ram while it simulated. over 200M voxels top. I really dont know if it is too much jajajajajajajajaj. One thing ive noticed is that I just cant render a single full hd frame in a reasonable time (mantra). Do you think it is because of the huge volume? Are the nodes properly set up? I have to deliver this for next week and I am thinking about swapping to 3dmax just for vray. 

Have a look if you have the time. I left 1 frame of the sim for you to take a look.

And again. Thank you a lot. The info you provided me helped me a lot to know the standars of a mid res sim

look.rar