Maya nParticles VS RealFlow in liquid simulations

[aghaanantyab]

realflow = good but maximum ram used

maya = does not require much memory but its bad

[ribenawrath]

this i great and all, but I have to say that from these demos realflow looks the most natural.

[michaelwbell]

Maya 2012 nParticle simulations are meant to be multi-threaded now, i believe? Should be interesting!

[bharatkumargupta]

while real flow is a seperate product and nparticles (maya) is built in, u can easily see the quality that maya tools offer in competition to stuff around , maya is amazing app

[4b10xs]

thanks for the insight.

[zappenduster66]

awesome comparison. could you do this for realflow 5 and maya 2011/2 too? =)

[1stKarkan]

I think realflow had a much better physic quality

[D4RK809]

You didn't mention If the times you posted where in hours or minutes or whatever, especially the first comparison is that 35 hours or minutes?

[ICPJuggalo1988]

@D4RK809 Sorry, yeah all the times are in hours:minutes:seconds, with hours only showing provided it took that long. All of these scenes were run over the course of one evening, so the longest test would be the second one with objects crushing particles, where Realflow took 3 and a half hours.

[blazerart]

hello, this is very useful to me. have you tried the latest maya? could you make a video similar to this but using the 2011 nparticles' dedicated liquid simulation system? I'm more curious if there's an improvement in accuracy by leveraging the new parameters.

Thanks.

[anim8ordude]

RealFlow wins hands down. Faster, more accurate, overall better result.

[supermix1337]

had 5 emitters and res at 200 and cpu usage around 50, i7 and 4 gb ram 2000mhz:d

[realflow100]

@supermix1337 thats freaky...

[supermix1337]

@realflow100 What's freeky? :)

[realflow100]

@supermix1337 only 50%CPU usage at that res and that much ram and processor speed. Thats whats freaky

[supermix1337]

@realflow100 Oh, 6 months ago I made that comment:D

[realflow100]

@supermix1337 lol :)

[realflow100]

@supermix1337 oh and i type too much now my hand hurts lol.

[Smitakus]

realflow the best

[Natorusama]

heh.. fail for maya. )

[gamrman55]

Ah lol, well I'm a lil new to Real Flow myself, but that's simple comparison to real life physics. I can tell you right now you have the visocity on Real Flow set a lil too high. Very simple and easily correctible. Now, the reason for the high cpu usage is because Real Flow was meant for powerful calculations and surprising accuracy for use in commercial industries. If you don't really know how to use it, then I suggest using Maya's thing there. Lol didn't mean to sound rude. All due respect. ^^

[joemoet2009]

ohh try to increase viscosity in maya or decrease it

[joemoet2009]

how do i make the sim time take a lot less?!! i can not upgrade my RAM!!!!

[joemoet2009]

ok but my computer is so messed up it turns itself off!!

[joemoet2009]

how do you cut a hole in a cube if you can?

[killmachine234]

@joemoet2009 I would use the cut tool to draw out a edge in the shape you want to cut, then delete the poly inside it on both sides then select the edges that you made on both sides and use the edge bridge tool.

[joemoet2009]

EMACHINES! = VERY BAD!my cpu goes 100% when simulating

[ICPJuggalo1988]

Having a CPU crank at 100% just means it's doing the hardest work it can, and is a good thing when you're asking for it. Having less than 1GB of ram could certainly be a bottleneck, these scenes were run with 4GB, but soon after I had to replace two sticks, and the drop even to 2GB was severe for simulations and rendering especially.

[realflow100]

@ICPJuggalo1988 hey you can easly get 16Gigs of ram for no money and no scrounging in the trash just changing settigns that are already on your computer and it is so easy right click on computer click properties go to advanced tab then click settings then go to advanced tab again then click change then uncheck system managed and manage all rives then click custom size then type 16000 into both boxes then YOU MUST HIT SET then click ok then restart your computer if it asks you. and there you go!!

[wasitrealy]

@realflow100 That's not RAM, that's Virtual RAM

[joemoet2009]

1 CORE

[joemoet2009]

of RAM INSTALLED!

[joemoet2009]

all i use is a plane and VASE!!!!!!!!!!!!!!!!!!!

[joemoet2009]

REAL FLOW i only have 895 MB OF RAM i just checked! WTF!

[joemoet2009]

the simulation goes very slow with 6000 particalles help whats wrong!!?

[ICPJuggalo1988]

Is it a simulation in RealFlow or Maya? Some other information, like how many cores and the speed of your CPU heavily factor in. Also, are there complicated high-poly objects that the particles collide with?

[anim8ordude]

@joemoet2009 I will guess that your computer is slow. If it's not, open the task manager to check that no other software is hogging anything, run a virus scan, get antivirus software if you don't have it and make sure that the amount of threads in the settings is the amount of processor cores you have, unless it's got hyperthreading technology.

[ICPJuggalo1988]

@anim8ordude

"I will guess that your computer is slow."

Based on...? Everything that is properly multi-threaded (RealFlow, Maya at render, or really any app 3D/compositing when rendering) ran just as expected, and relative to HP workstations that should provide the same or even slightly greater power my Q6600 setup tends to edge ahead in rendering/particle benchmarks.

Suppose if I drop a grand on an i7 I might get ahead a bit, but doubt the price/performance would be justified.

[realflow100]

@ICPJuggalo1988 it only has one core and its very slow but it is very well cooled with only one fan for the processor and a monster of a heatsink so the temp is fine i guess it's just plain slow but my new computer is much faster and dont get 100% used the whole time i use realflow and it is so much faster and stable but it gets pretty hot so i have to watch out but it rarely gets over 72C simulating realflow and it is idle at 45C

[realflow100]

@ICPJuggalo1988 oh and it is dual core

[planetmirth2001]

Very interesting - thanks for posting.

[SK82theTUNES]

Hey juggalo quick question,how small was ur particles? smaller then the default size or was it larger? and also great job with ur video man. im applying any things i see in it for some tests of my own.

[scionofwar]

Very nice scene! Not only does this show a great comparison between the two liquids, it also makes me feel happier about my investment with RealFlow.. I really have no need for Maya now since I use 3ds Max.

Awesome setups too! I loved the wave one, now I have a great idea on how to make my own ocean waves. Very useful!

The substeps on RealFlow seem to be too high almost for casual simulations, some people were saying to use only 50 substeps in some cases? That would be over 1/6 calc time!

[ICPJuggalo1988]

Yeah, lowering the substeps can certainly help reduce times, though I wonder if there's a way to check the average substeps. By default RealFlow assumes an adaptive substep of 3-333 times per frame, and when there is little happening it's obvious it uses far below it's limit, and ramps up as more collisions/calculations become necessary. Realflow and Max is a wicked combo, have fun with it!

[morgs2020]

I'd like to see a comparison of accuracy between Maya 2010 & Realflow given the same compute time.

[etudenc]

Also wanted to say "thank you!" for putting this together and posting your results. RF looks much more accurate in every case. Your splash setups were instructive, too! Cheers!

[ICPJuggalo1988]

Glad to help! I was curious and figured I wasn't the only one who wanted to see these two side by side.

[baban606]

Hi juggalo can u please upload a tutorial to create maya nparticles water effect creation.

[ICPJuggalo1988]

It's an almost insultingly easy setup, so I'll just type it here. First; you must be using Maya 2009 or later. Under the nDynamics menu set go to nParticles>and first select water as the type, then select create emitter. From there it's up to you to tweak the count, particle size, and various attributes to get the resulting particles to behave just how you want, but that's how to get nParticles to behave with liquid dynamics.

[baban606]

Thanks will give it a try and let u tell.

[Cadmium77]

This was a beautiful little video and it answered the very simple question I wanted answered perfectly

Thanks for doing the work to make it and thanks for posting it. Good stuff.

[dekekincaid]

I don't understand how one is more "accurate" then another? If one splashes more then another it simply means you need to change the settings of one to match another.

[ddsmooth]

good stuff, put alot of time into this. thanks

[ICPJuggalo1988]

No problem, I was really curious and for the most part and just set these up real quick and ran them overnight or while in class.

[nicorender]

I would like to see a comparison like this with Houdini Particle Fluids :)

[Jordanrom]

Interesting tests, thank you.

[LadyCobainTLK]

I wish I could have used realflow for my Underwater Nuke Explosion project in Tripp's class, but I was stuck with nParticles. I did get to make emitters with fish geo stand-ins though!

[organicoman]

Hi,

This is very interesting and very usefull, i was looking for this such test for a while. but buddy, if u can tell us what horse power have on your Computer(only for a coparisaon). but over all thank u for this test.

[ICPJuggalo1988]

Yeah, that makes sense. Even though Maya was 64-bit and RealFlow was only 32, I doubt it made a difference given the low memory consumption of the whole setup. Here's some relevant specs;

Quad-Core Q6600 at 2.7GHZ

4GB under XP Pro 64

[Spectrum3DYT]

64-bit is nothing like "just more than 4GB of RAM" like some people think.

It depends on how well the program was made, but it can be much faster than 32-bit calculations, so this comparison isn't really efficient.

But one thing is that you wrote that RF was multicored and Maya only with 1 so it's a bit strange, but while you didn't wrote the number of cores anyway, I've still noticed that in some cituations (in lite calculations), multi-C is slower than single for me, for some reason.

[Spectrum3DYT]

And another problem is that you've used a way too big particle size and probably also scene size (the realflow emitter has very large spaces) and it looks like floating blobs, so it's hard to compare the actually output quality.

[01DOGG01]

I haven't used either, but this nparticles thing looks like it's got too much damping on the motion.

Have u tried OE-Cake?

[ICPJuggalo1988]

Never heard of it, but from what I could see on the site's images, it looks pretty cool! It was brought to my attention that Maya defaults to a substep of 3 apparently, whereas RealFlow at default uses an adaptive sampling of up to 333 substeps, so that could account for both Maya's better sim speed and notably less accuracy.

[01DOGG01]

LOL that's a big difference.

You can't get OE-Cake from the site any more, but i have a download / installation video about it with links. You can find it in the oe-cake playlist on my profile page.

If you like playing with particles, it's definitely worth checking out. It's real time too. But you can create things which take hours to complete ("OE-Cake 1.1.0b - Nuclear Explosion" for example).

[brutalducky]

so bouncy

[GNUTM]

OHHH, TEACH ME.

[Currypl]

I think Real flow is the best by far. nParticles is not bad but RF is better in my opinion.

ans about diffrences in both samples it is all about settings. every slight change of some setting may change the animation completely.

[ICPJuggalo1988]

Too true. I tried to keep them generally level, but even minor adjustments would have profound changes, most evident in the wave simulation. Overall I can't deny the speed of Maya's Nparticles, even without proper CPU threading, but if you want anything legitimate, and remotely accurate, stick to RealFlow for now. I know I will!

[darkfox404]

RealFlow just seems splashier to the naked eye. RealFlows been in development for a bit. nParticles is a new feature hopefully Autodesk can refine it to be up to par with RealFlow.

I'm just hoping Autodesk doesn't destroy everything.

[ICPJuggalo1988]

"I'm just hoping Autodesk doesn't destroy everything. "

Right there with ya on that one. I'm already scoffing at new features, like the stupid bar in each window that has icons for the same options in the menu right above (at least that can be disabled), or Mental Ray being torn apart like in XSI...

[Cheesestraws]

To make an object be affected by nParticles you just need to turn it into an active nMesh.

[Cheesestraws]

It looks like as if the scale is wrong with nParticles. They are working on multi-threading nucleus, when multi-threading comes it should affect all of the nucleus due to the nature of the solver. Instances and goals have not been updated for nucleus yet so they still use the old pre-nucleus ones.

[Cheesestraws]

Also nParticles don't currently feature surface tension which might explain some of the differences you see, this is another feature that they are looking to add in a future release.

[mpingping1102]

I think Realflow better simulation and realistic than nParticle.

[l33yang2008]

nParticles is a new system, so I don't blame autodesk for being in accurate.

What they should've done is buy Realflow. They bought XSI, mudbox, and all these other ones. Might as well buy Realflow. Realflow is the winner for me. The only good thing about nparticles is it's particle collisions with one another. Other than that it's limited. Instances don't inherit the particles rotation values...which i think stinks for such an advanced module.

[KidderD]

fantastic tests. I also can not vouch for the validity, but it's certainly looks like you created a serious production test here. Now thread this sucker Autodesk!

[ICPJuggalo1988]

Especially given Maya was already in general faster having it properly multithreaded would really give it an edge. XSI's ICE engine is multi-threaded, I was disappointed to see nParticles still weren't. And this was in 64-bit Windows and Maya too.

[Terrox92s]

yay, finaly a new video