RAM usage by Real Flow

[mulls]

hi,
I am using Pentium Dual Core processor, with the power of 3.4Ghz, Graphics card - Quadro F/x 540 and and have 3 GB of RAM in my machine. Still the simulation process is taking the same time as it used to take when i was using 2GB of RAM. And the main thing is, Real Flow dosent even use 500MB of RAM.
So, my question was, is there any setting in Real Flow that will increase the RAM usage so that my simulation is faster?
Or does the simulation time depends upon something else, like processor or Graphics card?

Thanks in advance,

Mulls.

[ChristianZ]

Hi

in my experience RAM is not that much important with RealFlow. I could run a simulation with 500'000 particles with 1 Gb installed, but the processor has to calculate all of those particles. So I would say processor speed is the most important factor of all.

[tsn]

RF´s RAM management isn´t easy to see through. There are some easy things you could do to speed up your sim:

Preferences > General > RAM: Here you can select the amount of reserved RAM for RF. Though I don´t know exactly how this affects speed it´s recommended to use a rather high value. I think this acts more like a buffer for RAM usage peaks to disburden the hdd.

The next thing is to minimize the main window. Drag is as small as possible. This helps to decrease sim times up to 30%. Don´t choose perspective view while simulating. Up to 50% of the processor capacity is used for displaying the particles.

Use a fast hdd and store your results to an extra partition. This helps to avoid fragmentation effects.

Choose time steps carefully. It´s not always neccessary to run RF at 333 substeps, esp. for test sims.

Maybe someone else has more ideas (would be a nice thread to share those tricks)?

[stevester1]

tsn wrote: The next thing is to minimize the main window. Drag is as small as possible. This helps to decrease sim times up to 30%. Don´t choose perspective view while simulating. Up to 50% of the processor capacity is used for displaying the particles.

Better yet, run the command line.

Rule of thumb with realflow: when you're using RF, only work with RF. Give your pc a break, don't have background applications running.




As for the ram consumption itself, maybe your test scene doesn't require the 2 or 3 gigs of ram, maybe the simulation you're running only requires 500 MB. Crank the particles up and then tell us whether it is the same speed. And you have not mentionned if you are in 32 bit (with or without 3gb switch) or in 64 bit mode. The graphics card does not change your simulation speed...yet.

[shaun_michael]

Hey Mulls, please, there is no need to double post every question you have. It just makes the answers more difficult to find for other users.

I've moved Christian's post to this thread and deleted the other thread.

Thanks for your co-operation,
Shaun

[mulls]

hi,
thanx for your wonderful suggestions. I tried everything that was suggested. I am using 32 bit.
While trying these things to fasten my simulation, i found that, if not needed, we can reduce the max particles. This also speeds up the simulation. And offcourse killer daemons also helped me a lot, since the particles were moving all over the scene.

But I still cannot find this RAM setting (Preferences > General > RAM).


Thanks,
Mulls.

[DesignerV]

I usually use 80-100 max substeps in adaptive mode and press ald+d before running sim ;)

[tsn]

mulls wrote: But I still cannot find this RAM setting (Preferences > General > RAM).

Sorry, my mistake. It´s Prefernces > General > File Cache (MB).

[ChristianZ]

Yes, Alt+D disables the display and that should be almost as fast as the command line option, I guess.
Concerning RAM settings, the file cache probably has an effect on playing saved bin-files in the timeline, but not for the initial simulation (?).
The minimal substeps setting has a big impact on speed, however I had the idea that I see more crashes of simulations (especially in the rigid body setting) with lower substeps and that is something I would like to avoid, especially for simulations running for several days :glare:

[tsn]

ChristianZ wrote: Concerning RAM settings, the file cache probably has an effect on playing saved bin-files in the timeline, but not for the initial simulation (?).

Yes, ChristianZ, you´re right! So mulls, please just forget my comment above...