Hi,
Just tried to pour lately modelled/Blender 2.6/ pool w/water and noticed significant(x2) slow down,
also Speed coeff. does not seem to make a difference ?? Fluid sim is in everlasting development or what?
Hi
Which blender versions were you comparing?
I just tested baking a fluid simulation, and the results for the latest version of blender (current compiled version of 2.77) baked slightly faster than 2.69b using Ubuntu 14.04:
blender 2.69b: 1 hour 36 minutes versus
blender 2.77 : 1 hour 30 minutes.
I also tested the speed parameter for settings of 0.5, 1, and 2; and it appeared to be working as expected.
Although perhaps your setup is different. Maybe you could post your blend file (or a simplified version) at
http://www.pasteall.org/blend/
and I could compare my bake time to yours.
There was only a slight change to the fluid simulation in 2.77 -> the progress bar now displays percentage
complete, and if you hover your mouse cursor over the progress bar it displays time elapsed and time
remaining.
Hi bluecd
I found a few issues with your blend file; not sure if they’re causing the slow down, but they probably don’t help matters:
-
the ‘Volume Initialization’ for your pool obstacle is set to ‘volume’, and should be set to ‘shell’
(It’s best only to use volume when the object is “water-tight”, ie. has a well defined inside and outside (like a cube versus a plane). -
the pool obstacle contained ngons (not sure if this matters, but best to stick to quads, or tris)
-
the inflow object had no inflow velocity set
-
the scale was not applied on the Fluid domain and Inflow objects Ctrl+a -> Apply scale
Once these changes were made, the bake was slightly faster in 2.77a:
2.77a 25 minutes to bake (speed = 3.0)
~ 6 to 7 minutes to bake (speed = 1.0)
2.69b 28 minutes to bake (speed = 3.0)
~ 8 minutes to bake (speed = 1.0)
Note that at higher speed settings, baking time increases substantially, since the inflow produces more fluid.
If you bake at the 2 different speeds, you should see that at a speed of 1, the pool is filled to about 1/3 the height versus when a speed of 3 is used.
http://www.pasteall.org/blend/41458
Also, I forgot to ask what os you are using? Windows or …
I would be interested to know what your computer specs are and what your bake time is for the above file.
I was baking on an older quad core Q9450, and it seemed to be using all 4 cores at close to 100%.
hope that helps
Thank you for the research.
The file had been created some Blenders ago, so maybe this is partially the case…but:
- I set Inflow velocity only to earth const. accel and adjusted sim.speed in Domain’s settings instead, wrong.
- Applied Scale might also be the problem, always forget about it, my bad.
As is comes to specs: AMD Phenom X4 16GB Ram, Win10 64bit.
There must have been sth wrong w/this anyway because only a little fluid appears and thats it, to the end of sim.
Hi again
Just forgot to mention earlier that the inflow object’s volume initialization was set to volume, but it’s “open” on the end; which is probably the major culprit causing issues.
(So it would be best to add another face, or just use a plane and set the volume initialization to ‘Shell’)
Well, that obvious proof it works, however in my model it starts (speed already controlled) and after initially displaying some progress
simulation proceeds to the end but nothing more becomes visible.
Hmmm, that’s strange. So, just to be clear; you’re using the initial blend file that you posted? If you’ve made changes and are still having problems, maybe post an updated version and I’ll take a look at it.
I’ve had some strange occurrences like you’ve described, and they were usually caused by one of three things:
- Bad geometry - the fluid sim needs to be able to differentiate the inside versus the outside of your mesh (or the front versus back for shell like geometry). So things like mixing volume and shell geometry, inconsistent normals, internal faces will cause problems.
A quick check for non manifold geometry can find problem areas - switch to edit mode and press Ctrl+Alt+Shift+m -> any non manifold geometry will be selected.
Although I’m not 100% sure, but ngons may be an issue. I would avoid them or convert to quads, and ensure all normals are consistent.
- Incorrect setting of volume initialization,
and
- Setting the resolution to some extremely low value < 50 or so.
There are other things that can cause problems (negative scaling for instance), but the 3 above things are the most common.
Another thing that you might want to try, is to perhaps delete all the cached fluid files from your file browser for your current fluid simulation blend file to ensure that you’re starting with a blank slate and for whatever reason, there’s no cache files from a previous bake. I’ve never had problems with this recently, but I have heard other people mention it.