I'm having issues with my render time. I'm rendering a sequence and my first frame renders out at roughly 30 min, however as the frames progress they quickly become longer and longer to render out. Nothing changes in complexity in the scene as you can see below. Particle count for the sim is steady. I have a brand new windows 7 install with nothing but AVG running in the background. Ive restarted houdini and Windows several times and the result is the same. Starts off good then gradually builds up in time. As you can see 6 frames in Ive more than tripled from 26min to a whopping 1:33. Each frame in between increasing ~10-15min in length from the last. System is a brand new Ryzen 1700 @ a modest 3.8ghz OC with (2x16)32 Gigs of ram. Any ideas? Is this normal behavior?

My guess is it may be sim related. Do you have a sink? If not, then you may be adding more particles every frame and thus rendering them keeps using more time. How many particles are in your flip simulation?

Try placing a sink box at the bottom of your falls. Then at some point in the sim the particles count should start to level off. Move to that frame and render. The resulting time should be the expected average per frame.

Hard to say without seeing the scene or Mantra rendering settings. I would say you may need more Stochastic samples, try 9 or 12.

Overall I think the image looks lovely.

do you get the same behavior if you render without the waterfall?

are there potential spikes in the motion blur?

does the render time only increase or do some frames drop back down?

if you render the same later frame by itself is it still in the longer render time or does it render faster?

did you bake out the waterfall so it is only reading the data from disk or is it cooking the dops at the same time as the render? if it is, then it is possible you are beginning to swap which will increase render times as well…

Enivob
OK definitely did not have a sink for my mist particles but I'm getting the following error after selecting the smoke object and pressing enter to apply the sink. So, no luck so far getting that to work, any idea why?

Also, when I went into my whitewater sim, I've discovered the spray particles are falling through the main mesh used in the Autodop network as a static collision object. Not sure what to do about that one now either… Still new to FLIP fluids so not too sure where to go with this.

Btw, thanks for the compliment.

sidemjay
Havent tried without yet, but not having the sink leaves me to believe that's the culprit as further into the sequence now the mist is definitely gathering at the bottom of the waterfall.

Havent noticed any spike in motion blur.

Render time only increases

Havent checked re-rendering one of the later frames all on it's own yet…

Just the fluid is baked out. Haven't baked out the foam or mist, but I think I need to do that. Although when it is cooking a new frame, it doesn't seem like that time is included in the actual rendering of the image…

what is your ram usage doing as you render? it could be increasing to the point it puts the render into swap. the dopnets that are cooking during the sim will eat the ram you need to execute the render.

it is generally best practice to bake all sims to disk prior to rendering.


it is a mighty fine looking render I would love to see it in motion!

Thanks very much, I'll have to post the finished product all said and done! I'll check Ram usage on my next render, just trying to get that particle sink to work again tonight without any luck. I baked out the whitewater but I want to get the sink for the mist working before I bake it out, as it seems to be the culprit here for the increasing render times.

Any ideas as to why I get the above error when trying to apply the pryo to the sink? At a loss here…

This sink error is really frustrating.

My memory usage is maxed, even after having cancelled the render it does not drop…

In this last render pass while I was at work first frame was 20 min, eight frames later it took almost 3 hrs.

so it seriously sounds like this is an issue where your machine is going into swap and that is what is causing the increase in render time.

have you tried baking the sims out? that should help keep the render times to what they are when you start the renders

OK gotcha. Yeah I haven't baked the mist sim out yet which seems to be the most resource intensive of the three, yet I can't get the sink to work on eliminating particles at the bottom yet. Going to have to get that figured out tonight…

if you can provide a basic hipfile we could help you figure out the sink issue

OK will definity do, but I'm finding this mist sim is just totally eating my ram and bogging my system down. Without it I have 22 gigs of usage, with it I'm maxed and my view port grinds to a halt and everything takes forever to even just select. And I'm using a GTX 1080. When I first load Houdini and load the scene with mist the viewport is butter smooth, but when I start working things rapidly start to bog down. I baked the mist sim but even when I click Playback Simulation it still wants to cook out the sim and bog things down… Is 32 gigs just not enough?

Baking out and using Playback Simulation is too heavy - it has all the attributes to re-sim whilst all you want to do is render the final results.

What is the best way for caching the sim for rendering then? Still new to Houdini so I'm not sure how else to approach it.

So I disabled mist and get steady 20-25min/frame now no matter how far into the sequence. The mist sure does add a lot to the scene, however.

In this case I would take a 2-pass approach. Kick out a .bgeo.sc cache from the compression node inside of import_fluid. Then once that is generated, kick out a second pass from the surfacing node file cache. The surface cache is what you will render.

By kicking out the compression first you can simulate a higher quality fluid with the same ram.

Ah excellent, thanks! That is much better now for scrubbing though and obviously now rendering for the FLIP fluid and whitewater. Baking the mist out properly now, although the files are huge..

Hi Sneil,

I'm having the same problem with the spray component of the whitewater not properly colliding. Foam and Bubbles work perfectly.

I've tried 20 substeps and adding drag to spray but the problem is still there.

Any help would be greatly appreciated.