August 8, 2026 Yen Lily

FLIP Simulation Runs Out of Memory: Cutting Particle Count Without Losing the Look

Executive SummaryFLIP memory at a glance
Houdini 22

FLIP memory is driven by particle count, and particle count is driven by Particle Separation. Narrow Band, the right velocity transfer and separate caches cut memory without changing what the camera sees.

  • Particle Separation: halving it gives about eight times the particles; SideFX notes that increasing it lowers resolution and speeds up the sim
  • Narrow Band: removing particles deep inside the fluid reduces the memory footprint
  • Velocity transfer: APIC (Swirly) uses more memory than Splashy
  • Volume limits: Kill Outside Volume Limits stops stray particles from stretching the grid
  • Caching: one File Cache per stream; mesh and whitewater separately

A FLIP tank or ocean shot that runs out of memory usually has far more particles than the camera needs. Most of them sit deep inside the fluid, where nobody will ever see them, but each one still costs memory and time on every frame.

This guide shows which FLIP Object and FLIP Solver settings set the particle count, how Narrow Band and the volume limits reduce it, and how to cache and mesh so the final sequence does not run out of memory at the last step. Parameter names follow the Houdini 22 documentation.

Checked against SideFX Houdini 22 documentation for the FLIP Object, FLIP Solver, Narrow Band, fluid caching and Particle Fluid Surface, November 2026.

Particle Separation sets the budget

On the FLIP Object, Particle Separation sets the spacing between particles, and with it the resolution of the whole sim. SideFX notes that increasing it lowers the resolution and makes the simulation faster. Particle Radius Scale, Grid Scale and Collision Separation are all derived from it, so it is the first value to tune.

iRender
Particle Separation drives FLIP memory
PARTICLES
Particle Separation 0.1×1 particlesParticle Separation 0.05×8 particlesParticle Separation 0.025×64 particlesParticle count grows with 1 / separation³: halving the separation gives about 8 times the particles.
Relative particle countLikely to exhaust RAM
Arithmetic; SideFX FLIP Object docs

How to read the figure. Each bar shows how many particles the same tank needs at a different Particle Separation, relative to the first. At 0.05 it needs about eight times as many as at 0.1, and at 0.025 about sixty-four times, because particles fill a volume and the count grows with the cube of the resolution. The orange bar is where a workstation is likely to run out of RAM. Block out the motion at a coarse separation, then halve it once for the final only if the detail really needs it.

Narrow Band: keep particles where the camera looks

With Input Type set to Narrow Band on the FLIP Object and Enable Particle Narrow Band on the FLIP Solver, Houdini keeps particles only in a layer near the surface and represents the deep fluid as a volume. SideFX describes the benefit directly: removing particles deep inside the fluid reduces the memory footprint. Seeding a full volume wastes memory.

iRender
Narrow Band: fewer particles for the same surface
NARROW BAND
Full volume: particles all the way downNarrow Band: particles near the surface onlydeep fluid kept as a volumeBandwidth sets how deep the particle layer goes; attributes on removed deep particles are lost
Particles near the surfaceDeep particles removed by Narrow Band
From SideFX FLIP Solver and Narrow Band docs

How to read the figure. The left tank is a full-volume FLIP sim, with particles all the way down. The right tank uses Narrow Band: particles stay in a light-blue layer near the surface, and the dark area underneath is carried as a volume instead of particles. The surface the camera sees is the same, but the particle count is a fraction. The yellow note is the trade-off: Bandwidth decides how deep the layer goes, and attributes on deep particles are lost, so keep the band deep enough for splashes and churn.

Solver settings that quietly add memory

  • Velocity Transfer. APIC (Swirly) gives livelier swirls but uses more memory than the Splashy kernel. Keep Splashy unless the shot needs swirling detail.
  • Kill Outside Volume Limits and Dynamically Resize Fields. A few stray particles flying off can make the grid unwieldy. Killing particles outside the limits keeps the fields tight.
  • Use OpenCL. It accelerates the viscosity and pressure solves and is mostly useful for high-resolution sims with viscosity.
  • Substeps. SideFX notes that you usually will not need to increase substeps with FLIP, so do not raise them to fix memory or stability without a reason.

The settings in one place

The mockup collects the settings above, numbered in the order to check them.

iRender
FLIP settings that control memory
D1
FLIP ObjectInput TypeNarrow Band1Particle SeparationTest at 2× final2Particle Radius ScaleDefaultGrid ScaleDefaultFLIP Solver - Particle MotionEnable Particle Narrow Band3BandwidthJust deep enough for splashesVelocity TransferSplashy unless swirls are needed4APIC (Swirly) uses more memory than SplashyFLIP Solver - Volume LimitsKill Outside Volume Limits5Dynamically Resize FieldsStray particles make the grid unwieldyCachingParticlesFile Cache SOP (own stream)6SurfaceSeparate File CacheWhitewaterSeparate sim and cache
Setting to change
Schematic mockup, labels follow SideFX docs

How to read the figure. Badge 1 switches the FLIP Object to Narrow Band, and badge 2 is Particle Separation, the single biggest memory control. Badge 3 enables the particle narrow band on the solver, with a bandwidth just deep enough for splashes. Badge 4 keeps velocity transfer on Splashy unless you need APIC swirls. Badge 5 kills particles that leave the volume limits so the grid stays small. Badge 6 is caching: write particles to their own File Cache, then mesh and run whitewater from that cache rather than live.

Caching and meshing without running out at the end

SideFX warns that simulating in memory only might run out of memory. Use one File Cache per data stream, and cache the particles, the surface and the whitewater separately. Whitewater is its own simulation, driven by a Whitewater Source and Whitewater Solver reading the FLIP sim.

Meshing can fail even after the sim succeeds. On Particle Fluid Surface, raising Influence Scale greatly increases time and memory, and Partitioning should be enabled only if an OpenCL GPU runs out of memory. Convert To lets you output a Surface VDB or polygons. For storage, Fluid Compress shrinks caches but is lossy, so keep an uncompressed master if you may need to re-mesh.

Run through these before you start the final sim.

ChecklistBefore a final-resolution FLIP sim
0 / 8 done
Tip: Note the particle count and peak RAM of each test so you can predict the final before you halve the separation.

SaaS or IaaS for FLIP simulations

FLIP caches are large and sims must run start to finish on one machine, which affects where you can run them.

CompareNeutral guide

Where to run heavy FLIP sims

A neutral comparison of the two farm models for simulation work.

Your situationSaaS render farmIaaS render farm (iRender)
You need to simulate, not just renderMany farms only render; sim support variesRun the sim on the server as on your workstation
Your sim needs more RAM than you haveDepends on the farm's node specsEvery server has 256 GB ECC RAM
You need to mesh and run whitewaterUsually done locally, then uploadedMesh and simulate whitewater on the same server from the cache
LicensingProvided for supported tasksYou bring your own SideFX licences
Rule of thumb: if you only need to render meshed caches, SaaS can be simpler; if the simulation and meshing are the bottleneck, IaaS lets you run them in your own setup.

Running FLIP sims on an iRender server

iRender
How iRender works: your computer controls your own multi-GPU server
IaaS render farm
Your PC or MacNo powerful GPU neededRemote desktopLive screen1:1 like your PCYOUR OWN GPU SERVERbare-metal · not sharedGPUGPUGPUGPUGPUGPUGPUGPU1, 2, 4 or 8 × RTX 5090 (32 GB) or RTX 4090 (24 GB)Threadripper PRO · 256 GB ECC RAM · 2 TB NVMe SSDFull admin access · your setup is saved between sessions1Register, top upFirst $50+ top-up doubled2Upload your projectiRender Drive, free3Start and connectRemote desktop, any OS4Render, shut downBilling stops, per minute✓ Your workstation, just more GPUs: no queue to wait, no farm plugin, no job submission, no settings changed
Your mouse and keyboardServer screen streamed back
Schematic, not to scale

How to read the figure. The top row is the setup: your own PC or Mac connects by remote desktop to a whole bare-metal server with 1, 2, 4 or 8 GPUs, so you open your Houdini project there exactly as at home. The numbered row is the order of a session: register and top up, upload to iRender Drive before the server runs, start and connect, render a test frame and the job, then shut down so billing stops.

  1. Register an iRender account and top up; the first top-up of $50 or more gets a 100% bonus.
  2. Upload the .hip file and collision geometry to iRender Drive before you start the server.
  3. Start a server with 256 GB ECC RAM and connect by remote desktop.
  4. Install Houdini once with your own SideFX licence; the server image is kept between sessions.
  5. Simulate to disk with File Cache SOPs, then mesh, run whitewater and render from the caches.
  6. Download the caches or frames and shut the server down so billing stops.

Frequently asked questions

1. Why does my FLIP simulation run out of memory?

Usually because there are too many particles. Particle count grows with the cube of the resolution, so a small change to Particle Separation can multiply memory. Particles deep inside the fluid and stray particles outside the limits add more.

2. How does Particle Separation affect memory?

Halving it gives about eight times the particles. SideFX notes that increasing Particle Separation lowers the resolution and makes the simulation faster.

3. What does Narrow Band do?

It keeps particles only near the surface and carries the deep fluid as a volume. SideFX states that removing particles deep inside the fluid reduces the memory footprint; attributes on those particles are lost.

4. Is APIC worth the extra memory?

APIC (Swirly) uses more memory than the Splashy kernel. Use it when the shot needs swirling detail; otherwise Splashy is cheaper.

5. Should I cache the mesh and whitewater separately?

Yes. SideFX recommends one File Cache per data stream. Whitewater is its own simulation driven by the FLIP sim, so cache the particles first and build the rest from that cache.

6. Can I run FLIP sims on iRender?

Yes. Install Houdini on the server with your own SideFX licence and simulate on a machine with 256 GB ECC RAM; the server image is kept between sessions.

Run your Houdini scenes on iRender

iRender is an IaaS GPU render farm: you rent a bare-metal server with 4x NVIDIA RTX 5090 (32 GB) GPUs, an AMD Ryzen™ Threadripper™ PRO CPU, 256 GB ECC RAM and a 2 TB NVMe SSD, connect by remote desktop and work on it like your own workstation, with full administrator access. You install Houdini and use your own SideFX license; your server image is saved between sessions, so Houdini, your packages and caches stay installed. For Karma XPU and GPU-accelerated solvers we recommend the 4-GPU servers.

Multi-GPU servers are billed per minute (1-GPU servers have a one-hour minimum), and sessions of 3 hours or more get 10% off. Our 4-GPU servers are liquid-cooled; in our own measurements the GPUs hold 60–70°C under sustained render load. Elite support is available 7 days a week by hotline, live chat, Skype, WhatsApp and email.

This video walks through rendering a Houdini Karma XPU scene on an iRender server in five steps:

New users get a 100% bonus on the first top-up (minimum $50). See the Karma XPU render farm page for server options, or download the iRender GPU app.

If Karma XPU renders are slow on your machine, iRender’s Karma XPU render farm gives you a dedicated remote server where you install Houdini, bring your own license and render on multiple RTX 5090 GPUs (for Karma XPU, 4 GPUs is usually the most efficient setup).

Accelerate your production pipeline with high-performance IaaS render farm services and dedicated bare-metal workstations powered by 4x NVIDIA GeForce RTX 5090 (32GB GDDR7), high-frequency AMD Ryzen™ Threadripper™ PRO processors, 256GB ECC RAM, and dedicated 2TB NVMe SSD storage tailored for your pipeline:
, , , , , , ,

Yen Lily

Yen Lily is a customer support specialist at iRender who helps 3D artists set up Cinema 4D, Redshift, Maya and Houdini projects on iRender's multi-GPU servers. She has written more than 200 guides for the iRender blog, most of them on Redshift and GPU rendering workflows.
Contact

iRENDER GPU

Render Farm Pricing
Quick Start
Multi-GPU Benchmarks
iRender GPU App
Blog
Documentation
The Render Farm
The Company
Terms & Conditions

INTEGRATIONS

Autodesk Maya
Blender
Cinema 4D
Houdini
Karma XPU
Lumion
KeyShot
Unreal Engine
Twinmotion
Redshift
Octane
And many more…

iRENDER TEAM

Elite support 7 days a week
SATURDAY – SUNDAY: 6:00 AM – 11:59 PM
(UTC+7)
Hotline: (+84) 912-785-500
Skype: iRender Support
Email: [email protected]
Address 1: 68 Circular Road #02-01, 049422, Singapore.
Address 2: No.22 Thanh Cong Street, Hanoi, Vietnam.

Contact
이 페이지는 한국어로도 제공됩니다. 한국어로 보기 →