Games & Commercial Applications

This document answers the product questions, as distinct from the engineering ones:

• Where does MPM actually ship today?
• Has any commercial game used it — let alone built its core mechanic on it?
• How novel would a real-time, MPM-native game be, and how should that be talked about?

MPM is well-established in offline film visual effects (VFX) and engineering simulation, increasingly seen in real-time tech demos and research, but — to the best available knowledge — no shipped commercial game uses MPM as its core physics, which is what makes an MPM-native game plausibly unprecedented.

1. Where MPM actually ships today

MPM is not exotic — it is production-proven — but almost entirely outside of shipped games. The landscape breaks into five tiers, ordered by how battle-tested each is. The commercial-DCC tier is the newest of them, dating from Houdini 20.5 in July 2024:

• Offline film VFX — the most mature use. MPM’s graphics debut was Disney’s 2013 snow paper, and its in-house engine Matterhorn powered the snow in Frozen and is integrated into the production pipeline for Big Hero 6 and Zootopia (stated in the SIGGRAPH 2016 course abstract). This is offline, render-farm simulation — minutes-to-hours per frame — not real-time.
• Engineering simulation — mature, decades deep. MPM’s original home (Sulsky et al. 1994) is computational mechanics: geotechnical engineering (landslides, slope stability, soil–structure interaction), high-velocity impact, and granular/large-strain problems. Commercial and open-source engineering solvers exist (CB-Geo MPM, Anura3D). Again: accuracy-first, not real-time.
• Commercial DCC tooling — new since 2024, and it moves the baseline. SideFX Houdini 20.5 (July 2024) ships a GPU-accelerated MPM solver as a stock feature — written in OpenCL, presented to artists as the FLIP solver extended to solid mechanics, with snow, mud, soil, concrete, metal, jello, rubber, honey and sand presets, multi-material interaction, and rigid coupling. It is offline VFX, not real-time. But it means MPM is no longer “a method a studio implements”: it is a checkbox in the tool most VFX shops already own. cebas thinkingParticles exposes MPM as artist-facing operators too. So MPM is no longer exotic in commercial software, even though it remains rare in games — the two statements are about different products. Links in References §3.
• Research & tech demos — active, real-time emerging. Graphics research keeps extending MPM (CD-MPM dynamic fracture, AnisoMPM, CRESSim-MPM for surgical soft-body). Real-time MPM tech demos in Unity and WebGPU are now common — but a demo is not a shipped product.
• Shipped commercial games — effectively none (as core physics). See §2.

2. Is MPM in commercial games?

MPM does reach shipped games, but through a narrow door. No well-documented commercial game runs MPM as its core physics; what does ship is MPM as a liquids effects layer, supplied by middleware rather than written by the studio. The picture below is drawn from general web search, GDC (Game Developers Conference) and SIGGRAPH talk records, Steam, devlog channels, vendor case studies and the EA SEED repository.

Misattributions — games people assume use MPM, but don’t

A surprising number of “physics-y” games get mentally filed under MPM. None of these actually use it:

Two more come up often enough to be worth settling: NetEase’s Where Winds Meet and Naraka: Bladepoint have no MPM claim attached to them, and the GDC vault carries no production MPM talk.

The pattern: granular/fluid look ≠ MPM method. Cellular automata and voxels are far cheaper and have shipped widely; MPM’s continuum accuracy has not been the tool studios reach for.

The closest real things

Five cases sit closest to a game built on MPM: two games, the middleware that puts MPM into other studios’ games, a prototype that was abandoned, and a company that holds the patent without a disclosed shipment.

• Liquid Crystal (Grant Kot / kotsoft) — an in-development indie MPM sandbox game with publicly playable pre-alpha demos, built on Kot’s own MLS-MPM-lineage engine. It is not commercially shipped, its internals are unpublished, and the player character appears kinematic — but it is an announced, playable MPM-native game project.

  • The design is all-continuum, not an effects layer. Per the author, “everything is simulated with particles,” including rigid-reading structures that fracture and shatter; the stated design is “a physics sandbox built around a unified particle-based physics model. Every material is configured from the same set of parameters.”
  • What ships. A Temples level demo and a Materials sandbox, as DirectX 12 / Windows builds using mesh shaders for particle rendering and up to four split-screen views, in 0.5M- and 1M-particle variants with separate SSE2 and AVX2 CPU builds. Public performance reports span ~30 fps on an RTX 2070 SUPER down to single digits on weaker machines, so it is a demo rather than a tuned product.
  • The configuration, from the author. Explicit time integration at a single step per frame, a 512×512×128 grid, particle resolution adaptive with about eight per cell at the surface, and per-material feature toggles so a liquid does not pay to track a deformation gradient it never reads. The physics runs on the CPU across all cores, leaving the GPU free for ray tracing — the exact inverse of the GPU-resident assumption most MPM literature makes. Three of those are work-elimination rather than throughput tuning, and they are what lets the domain reach that size at all.
  • It has hit the multi-field problem. Kot has publicly described needing “to put solids on a separate velocity field from liquids so the particle velocities don’t blend together” — the same single-velocity-field limitation that multimaterial contact MPM (Theory §5) exists to solve.
  • His two decades of MPM work, the port tree it seeded, and the part of his catalog that is not MPM are in References §3. Demos: Liquid Crystal Demo 2024 · Temples · Materials · the web MPM demo (lc1)

• Zibra Liquids (ZibraAI) — commercial Unity/Unreal middleware for real-time fluids built on MLS-MPM (atomics + radix-sort P2G), shipping in games since ~2021–22. The method is confirmed by the vendor: Zibra’s Approaches to real-time fluid simulation in visual effects names MLS-MPM as the method “from which our tool, Zibra Liquid, was born,” and describes its APIC basis. Because it is an off-the-shelf asset, every title shipping Zibra Liquid ships MLS-MPM — so MPM in games is not one product but an unknown-sized tail of them, with no public customer list. It is a liquids effects layer rather than core physics. zibra.ai

• Echoes of Somewhere (Jussi-Petteri Kemppainen) — a 2.5D point-and-click adventure whose first episode’s story is about water, using Zibra Liquid — and therefore MLS-MPM — to build game mechanics based on liquid control rather than an effects layer. A free demo was published on itch.io in December 2023; the Steam page for Series 1 reads “Coming soon”, so a playable public build exists and the full game does not. site · Zibra case study

• Hydrolab (Lee Vermeulen, co-founder of Alientrap Games) — an AR/VR liquid-puzzle game on Zibra Liquid where the puzzles are the liquid. It was a 2024 prototype and was abandoned in favor of voxel projects, so it never shipped. Zibra case study

• Tencent — holds two MPM patents through Tencent America LLC, both naming authors it employed at its Game AI Research Center: US11966999B2, Real-time simulation using material point method on GPUs, whose independent claims all require multiple GPUs; and US11373353B1, the counterpart of ASFLIP (Fei et al. 2021). Both are rows in Patents §3. Its disclosed shipped work is AI systems (QQ Speed Mobile, PUBG Mobile) rather than MPM, and Chinese-language sources are thinly covered in English indexes, so a Tencent MPM shipment can be neither confirmed nor ruled out from the public record.

3. What this does not settle

The survey covers what is discoverable, which is not the same as what exists. Two limits are worth keeping in view:

1. Studios rarely publish solver internals. A small or mid-sized studio could have shipped MPM-based effects without documenting or marketing the method, and nothing in the public record would show it.
2. “Core physics” and “a localized effect” are different things. A game built around MPM as a central mechanic has no public precedent. MPM used quietly for a localized effect — a snow, mud or goo patch in one level — is far more plausible to exist already somewhere unindexed. The two are easy to conflate and describe very different games.

4. Why no shipped game uses MPM extensively

The absence is not an accident — there are concrete structural reasons:

• There is no general-purpose off-the-shelf MPM tool. Unlike rigid-body physics (PhysX, Havok, Box2D), MPM has no drop-in general engine middleware — the one commercial offering (Zibra Liquids, §2) is fluids-only, an effects layer. A studio wanting MPM as core physics must build the solver itself.
• Unity does not ship MPM. Unity’s built-in physics is rigid-body (PhysX in 3D, Box2D in 2D). Any MPM in Unity is custom code a developer writes on top of the engine — there is no “Unity offers MPM but nobody uses it” paradox; the engine never offered it. (Same for Unreal, Godot, etc.)
• The real-time barrier is real. A studio building its own solver inherits MPM’s limitations (see Theory §7): the small-timestep stability limit (the Courant–Friedrichs–Lewy, or CFL, condition), high memory cost, and the particle-to-grid (P2G) scatter parallelization bottleneck. Hitting interactive frame rates on consumer hardware was, until recently, very hard.
• The risk/reward hasn’t favored it. Cellular automata and voxels deliver “destructible/granular” feel at a fraction of the cost and complexity, so the commercial incentive to take on MPM’s R&D burden has been weak.

4.1 What ships instead — the incumbent water & snow stack

The last bullet deserves specifics, because “what a studio would use instead” is the actual competition, and it is very good and very cheap. Two techniques own the space, and neither is a continuum solver:

Water → 2D shallow-water height fields. Not cellular automata — a real physics solver, just a 2.5D one (full method in Related Methods §1.5). The reference implementation is Chentanez & Müller’s Real-time Simulation of Large Bodies of Water with Small Scale Details (SCA 2010): shallow water equations on a height field, with particles spawned where the height field cannot cope (breaking waves, waterfalls, splashes) and reabsorbed on landing. The cost is startling — complete scenes at ~4–18 ms/frame on a 2010-era GTX 480.

That paper is not merely academic: Fluid Flux (Imaginary Blend / Krystian Komisarek), a commercial Unreal Engine water system released April 2022, is a direct implementation of it — the developer said so explicitly when asked, linking the paper. Its shipped envelope is a useful benchmark for what “good enough water” costs today:

It is also externally validated as physics, not just as an effect: a 2025 peer-reviewed digital twin of a river-experiment facility used Fluid Flux as its solver and compared the results against a real-scale flow experiment. A game engine’s water plugin standing up to hydraulic-engineering scrutiny is a fair measure of how mature this incumbent is.

Layered simulation is the shipped pattern. An indie voxel engine handling oceans will typically run a bulk voxel fluid sim for the water body, wave and tide motion applied as oscillating “target height” masks, a connectivity flag marking which water is joined to the ocean (so rock pools disconnect as the tide drops and keep their water), and a separate, higher-resolution, surface-only 2D ripple simulation running near the player for wakes and interaction. Nobody runs one solver at one resolution — the budget goes where the camera is. That is the same demote/promote logic the height-field papers use, arrived at independently.

Snow, sand and mud → displacement textures + parallax occlusion mapping. Deformable ground in shipped games is overwhelmingly not simulated at all (full method in Related Methods §2.3): a mask written by simple analytic primitives at the feet, a diffusion pass to soften it, and a parallax shader that marches the view ray through a height texture so the surface reads as displaced without any geometry. Indie implementations of exactly this run comfortably in real time on laptop hardware, and are genuinely convincing to look at.

The critique of them is the more useful artifact, because it is a public catalog of where the illusion breaks, written by players rather than researchers: no compaction (a repeatedly-walked path never packs down into a faster route), no cohesion (the thin pillar of undisturbed snow that survives between a walker’s legs cannot exist), no conserved volume (displaced material doesn’t have to go anywhere), no bulk failure (no post-holing, no collapse), and — most tellingly — deep examples read as sand, flour, or mud rather than snow, with repeated objection to bodies moving through waist-deep material unimpeded.

Where the two approaches actually differ. For shallow decorative deformation or a plain river, the incumbents look excellent at a fraction of the cost, and remain the sensible engineering choice. What they do not produce is the list above: compaction, cohesion, conserved volume, bulk failure, depth-correct resistance, material mixing, and phase change. Those are consequences of a constitutive model rather than features that can be bolted onto a height field or a displacement texture, so refining the incumbent stack does not converge on them. The trade is straightforward: MPM produces those behaviors and costs considerably more per frame.

5. What makes MPM-native games newly viable

The barriers in §4 are eroding — which is precisely why now is an interesting moment for an MPM-native game. Three developments converge (all detailed in Theory and Implementation):

• MLS-MPM (Moving Least Squares MPM, 2018) makes each step ~2× cheaper and drops the shape-function gradient — modern real-time work starts here.
• PB-MPM (position-based MPM, 2024), from EA’s research lab SEED, makes MPM stable at any timestep, explicitly framed for “real-time applications where the hardest requirement is stability under autonomous, potentially violent player input” — i.e. a game.
• GPU optimization (fixed-point atomics, sparse grids, spatial sorting) brings the particle counts and frame rates into interactive range.

The gap an MPM-native game would step into is exactly this: the enabling techniques are published, and their reference implementations are permissively licensed (MIT / BSD-3), but no shipped game has yet committed to MPM as its core. A permissive license governs copying that source and nothing more — the underlying techniques carry their own patents, which is a separate question with its own document (Patents).

The frontier keeps moving (2024–2026): XPBI (SIGGRAPH Asia 2024 — real-time continuum inelasticity in a PBD pipeline), CK-MPM (SIGGRAPH 2025 — compact-kernel performance), convex MPM↔︎rigid coupling pushed to interactive rates on GPU (Zong et al. 2024; Yu et al. 2025, the Drake/robotics line), and PB-MPM derivatives (e.g. an affine-projection stabilizer, The Visual Computer 2025). All still research and demos — none is a shipped game — but the field is moving quickly. Full citations in References.

6. EA SEED, and where PB-MPM came from

SEED (Search for Extraordinary Experiences Division) is EA’s internal applied-research lab, and the group that published PB-MPM:

• Launched at E3 2017; sits inside EA Worldwide Studios (offices in Stockholm and Los Angeles); led by Colin Barré-Brisebois.
• SEED does not ship games itself — it turns experimental ideas into techniques used across EA titles, and publishes openly so others can build on the work.
• That open-publication model is why PB-MPM is public, demo-only and BSD-licensed rather than an internal technique.
• Its other work gives the measure of the lab: Project PICA PICA (real-time ray tracing), Swish (neural-network cloth simulation in Madden NFL 21), Voice2Face (audio-driven facial animation), reinforcement-learning goalkeepers in EA SPORTS FC 26, collusion detection in team-based multiplayer, and Vaporwave, a scalable real-time smoke solver. The full list is on SEED’s publications page.

Sources for this document

• Games context: assembled from general web search, GDC/SIGGRAPH talk records, Steam, devlog channels, vendor case studies and the EA SEED repository.
• Shipped-destruction context (how games actually do it — falling-sand, voxel, heightfield baking): Purho, Exploring the Tech and Design of Noita, GDC 2019 — gdcvault.com/play/1025695; Brown, How Games Do Destruction, GMTK, 2025 — gmtk.substack.com; From Dust (2011) as the shipped sim-material→static-terrain precedent (lava hardening into walkable terrain).
• §4.1 incumbent stack: Chentanez, N. & Müller, M. Real-time Simulation of Large Bodies of Water with Small Scale Details, SCA 2010 — PDF (source of the method and the GTX 480 timings). Fluid Flux — product page & documentation (feature set, resolution/coverage limits, GPU benchmarks, engine support), demo build; the developer’s attribution of the algorithm to the 2010 paper and the ~0.5–0.7 ms / RTX 2060 figure are the developer’s own, stated at the 2022 reveal. External validation: A Digital Twin of River Experiment Infrastructure Based on a 3D Game Engine and Validation of Water Flow with a Real-Scale Experiment, Applied Sciences 15(23):12507, 2025 (flow reproduction validated at the Andong River Experiment Center) — mdpi.com/2076-3417/15/23/12507 · DOI: 10.3390/app152312507. The displacement-snow technique notes (mask primitives, diffusion, normal smoothing, POM) and the list of what players say it cannot do are drawn from indie devlogs and their reaction threads rather than from published work.
• Disney production use: SIGGRAPH 2016 MPM course notes (abstract names Frozen, Big Hero 6, Zootopia) — PDF; Stomakhin et al. 2013 snow paper — Disney Research.
• EA SEED / PB-MPM: github.com/electronicarts/pbmpm · SEED · SEED 10-year anniversary.
• Real-time enablers (MLS-MPM, PB-MPM, GPU): see Theory and Implementation; full citations in References.
• Engineering MPM solvers: CB-Geo mpm, Anura3D.