Some Turbulencefd's issues

Workflow
1

I always love Turbulencefd, but it still have some issues that long time no fixed.

  1. Cannot convert .bcf file that bigger than 4.1 GB with 1 channel ( like temperature or density … ) , ( and 4.1 * 3 = 12.3 GB with 3 channels ) ( and 4.1 * 3 + 4.1 * 3 = 24.6 GB with 3 channels + 1 channel velocity including 3 dimension x,y,z )

  2. Too slow ( not opimized ) vdb convert process. I take all days to convert a bcf sequence. Even longer than times to simulate that sequence. Example, if I have 300 bcf. frame in folder named cache 001 ( the final frame is a 4GB bcf file, pretty heavy ). It will takes forere ( I mean it’s really forever ) to convert 300 frame. I used to wait for 2 days and it’s still running. So have to do this steps:

  • Step 1: Copy 30 bcf. file to a new folder, convert bcf2vdb in that folder, delete 20 bcf. files
  • Step 2 : Copy 20 bcf. file to a new folder, convert bcf2vdb in that folder, delete 20 bcf. files
  • Step 3 : Copy 10 bcf. file to a new folder, convert bcf2vdb in that folder, delete 10 bcf. files
  • Step 4 : Copy 5 bcf. file to a new folder, convert bcf2vdb in that folder, delete 5 bcf. files
    … until I convert all of 300 files. and It take a morning to finish.

I have feeling that this issue due to the convert tool does not convert frame by frame, but it covert many frames at the same time. This lead to crash. I wish it have better optimized workflow.

2

I don’t see bcf2vdb getting any updates as reactions has this issue solved…at least for smaller rea files

1 Like
4

If you are rendering with Arnold you do not really need to convert to vdb…even RS is OK with bcf but slower

5

With large simulate, converting is really need because Tfd viewport perfomace is weak. Also, the vdb is needed to use Arnold complex volume shading system.

1 Like
6

Agreed the live updating of the IPR is a bonus but it works fine for me with bcf. But for shading ? I can use volume sampling in the shader network editor using a TFD container with bcf. The float nodes are simply looking for a channel name and are not concerned about vdb or bcf

1 Like
7

Yes. You can use step size for vdb to improve rendertime much faster. I use it for multiscattering with many bounce effeciently.

8

You can use step size on a tfd container with an arnold tag