Pixar, NVIDIA and SideFX: 30 Years of Rendering Technology

1980-1995: The Foundations

The history of VFX rendering technology is one of constant evolution, breakthrough innovation, and a gradual convergence between competing industrial players. Three companies in particular - Pixar, NVIDIA and SideFX - have shaped the technological landscape of modern 3D rendering and visual effects, not as isolated inventors but as an interconnected ecosystem where innovation, acquisition and partnership progressively produced shared standards and interoperable workflows.

The Genesis of Pixar and RenderMan

Pixar's story begins inside Lucasfilm. In 1979, George Lucas recruited Ed Catmull to lead Lucasfilm's new Computer Division, tasked with bringing digital technology into filmmaking. The division gathered exceptional talent including Alvy Ray Smith, Loren Carpenter and Rob Cook. In 1984, the team developed the REYES algorithm ("Renders Everything You Ever Saw"), an innovative rendering approach using "micropolygons" to subdivide complex surfaces into small, efficiently renderable elements while supporting advanced lighting effects - it would later become the core of RenderMan.

In 1986, Steve Jobs bought Lucasfilm's Computer Division for $10 million and formed Pixar as an independent company. Under this new banner, the team continued developing the rendering technology that officially became RenderMan in 1988 - introducing a standardized programming interface (the RenderMan Interface), a shading language for mathematically describing surface properties, the ability to handle unprecedented scene complexity, and advanced techniques for simulating complex lighting phenomena. RenderMan was among the first commercially available render engines built specifically for cinematic visual effects, establishing concepts that became industry standards.

Mental Images and Mental Ray

In parallel, a German company named Mental Images was founded in Berlin in 1986 by physicists and computer scientists including Rolf Herken, immediately developing Mental Ray, a ray-tracing-based render engine. Unlike REYES's micropolygon approach, Mental Ray simulated the physical path of light rays through a virtual scene, naturally producing optical effects like reflections, refraction and shadows with high physical accuracy. From its 1989 commercial release, it was distinguished by physically accurate rendering, an extensible architecture via programmable shaders, efficient distributed rendering across multiple machines, and optimized memory use for very complex scenes - quickly establishing Mental Images as a serious competitor.

SideFX and the Beginnings of Proceduralism

In 1987, Kim Davidson and Greg Hermanovic founded Side Effects Software (later SideFX) in Toronto. Their first product, PRISMS, took a fundamentally different approach to 3D creation, focused on procedural generation rather than direct modeling - built on connectable "ops" (operators) forming data-processing networks, where upstream changes automatically propagated through the whole system. Though PRISMS initially relied on third-party render engines, this procedural approach laid the philosophical foundation that would always set SideFX apart: 3D creation should be dynamic, adaptive and parametric rather than static.

NVIDIA and the Dawn of Graphics Acceleration

In 1993, Jensen Huang, Chris Malachowsky and Curtis Priem founded NVIDIA to build specialized graphics processors, at a time when graphics processing was mostly CPU-driven with limited help from simple graphics circuits. In 1995, the company launched the NV1, its first GPU, integrating 2D/3D graphics and audio processing - a commercially mixed result, but the start of an acceleration era that would transform VFX in the following decades.

Toy Story: Commercial Validation

In 1995, Pixar released Toy Story, the first feature-length film entirely rendered in computer graphics, using RenderMan. Its critical and commercial success validated Pixar's business model and set a new bar for animation and visual effects, encouraging other studios to invest heavily in 3D rendering technology and accelerating adoption industry-wide.

1996-2007: Maturity and Specialization

Houdini and the Procedural Revolution

In 1996, SideFX launched Houdini, PRISMS's successor, refining the procedural approach with an even more sophisticated node system for modeling, animating and simulating physical phenomena non-destructively. Its fully procedural architecture, advanced physical simulation (fluids, destruction), expression/function system controlling every aspect of a scene, and plugin/custom-code extensibility quickly made it the tool of choice for complex effects - explosions, fluids, crowds, destruction - establishing SideFX as a major force in studio VFX pipelines.

Mantra

In 2001, SideFX introduced Mantra, its own render engine natively integrated into Houdini. Inspired by RenderMan, Mantra also used a micropolygon-based approach but with deep integration into Houdini's procedural system, supporting Micropolygon Rendering, raytracing, Physically Based Rendering, and volume rendering for smoke, clouds and other volumetric phenomena. Its major advantage was seamless handling of Houdini's own procedural data and dynamic simulations - geometry that would be difficult to export cleanly to third-party renderers.

NVIDIA's GPU Generations and CUDA

Between 1997 and 2006, NVIDIA shipped successive GPU generations: the RIVA 128 (1997, one of the first capable consumer 3D accelerators), the GeForce 256 (1999, which NVIDIA itself branded the first "GPU"), the GeForce 3 (2001, introducing programmable shaders), and the GeForce 8 Series (2006, the first unified CUDA architecture). In 2006, NVIDIA introduced CUDA (Compute Unified Device Architecture), a parallel computing platform letting developers use GPUs for general-purpose, non-graphics computation - a foundational shift turning GPUs from simple graphics accelerators into general-purpose parallel processors, eventually accelerating offline rendering, physics simulation, and later AI.

NVIDIA Acquires Mental Images

In 2007, NVIDIA acquired Mental Images, the company behind Mental Ray - a strategic move marking NVIDIA's official entry into professional rendering technology, gaining a mature, widely adopted rendering technology, an expert development team, established relationships with major 3D software vendors, and a foundation for GPU-accelerated rendering. This acquisition laid the groundwork for Iray, a GPU-optimized render engine.

Early Standardization

This period also saw early moves toward openness: in 2000, NVIDIA and ATI (now AMD) standardized shader programming interfaces; in 2002, Pixar began sharing parts of its RenderMan technology with the community; in 2003, Open Shading Language (OSL) emerged as an open-source alternative to proprietary shading languages - early signs that standardization and collaboration could benefit the whole industry.

2008-2016: Convergence and Standardization

The Rise of GPU Rendering

As GPU power grew alongside CUDA, a new generation of GPU-optimized render engines emerged: NVIDIA's OptiX (2008, a GPU-accelerated raytracing framework), Iray (2010, built on Mental Ray technology but optimized for GPUs), and Cycles for Blender (2011, developed by Brecht van Lommel, supporting CUDA GPU rendering). These marked a paradigm shift - GPUs proving capable of significantly outperforming CPUs for certain rendering calculations, particularly path tracing.

Pixar Opens Its Technology

Between 2008 and 2016, Pixar made several strategic moves to open proprietary technology: releasing parts of the RenderMan Interface Specification (2009), open-sourcing OpenSubdiv (2012), and - most significantly - making USD (Universal Scene Description) open source in 2016. USD was designed as a file format and toolset for exchanging and composing complex 3D scenes across different software and pipelines, offering a flexible data model for complex scenes, a composition system for assembling and modifying scene elements, a referencing mechanism for reusable assets, and an extensible framework for new data types.

Hydra

In 2014-2015, Pixar developed Hydra, a highly flexible rendering framework built into USD, introducing the concept of "render delegates" - letting different render engines plug into the USD ecosystem through a common interface. Hydra abstracts scene description from render engines, lets multiple renderers integrate into the same pipeline, and provides an extensible architecture supporting interactive real-time visualization alongside offline rendering - letting technologies like OpenGL rasterization and raytracing coexist in one pipeline.

Cross-Pollination: Brecht van Lommel

An illustrative episode of industry knowledge exchange: Brecht van Lommel developed Cycles for Blender in 2011, left the Blender Foundation for Solid Angle (Arnold's creator) in 2014, then returned to Blender full-time in 2018 - a path showing how expertise circulates between projects and companies even where no code is shared.

Industry Consolidation

In 2016, Autodesk acquired Solid Angle, the company behind Arnold (developed by Marcos Fajardo), an unbiased path tracer that had become one of the most widely used render engines in VFX and animation - part of a broader consolidation trend of major software vendors integrating advanced rendering technology directly into their suites. In the second half of the 2010s, NVIDIA began developing its RTX technology for hardware-accelerated real-time raytracing, building dedicated raytracing hardware units, denoising algorithms for low-sample-count quality, and the APIs to expose them - groundwork that would surface publicly in 2018.

2017-Today: The Era of Interoperability

Since its 2016 open-source release, USD saw rapid, broad adoption: initial uptake by major animation/VFX studios (2017-2018), major 3D software vendors announcing support (2019), integration into game engines like Unreal Engine and Unity (2020-2021), and the formation of the Alliance for OpenUSD in 2023 to formally standardize and evolve it. This adoption transformed production pipelines, enabling unprecedented interoperability and cross-tool collaboration.

In 2018, NVIDIA launched its RTX 20 series GPUs on the Turing architecture, introducing dedicated RT Cores for hardware-accelerated raytracing - for the first time making raytracing accessible in real time for consumer applications, letting games integrate raytraced reflections, shadows and global illumination, letting architectural/product visualization offer interactive photoreal previews, and letting VFX artists preview near-final-quality lighting. Subsequent RTX generations (30 and 40 series) kept narrowing the gap between real-time and offline rendering quality.

In response to USD and Hydra's rise, SideFX built two major technologies: Solaris, introduced in Houdini 18 (2019) as a native-USD environment for layout, lookdev and lighting - creating and editing USD scenes directly in Houdini, assembling complex scenes from multiple sources, non-destructively editing existing scenes via USD layers, and integrating with Houdini's procedural system for USD asset generation - and Karma, a new render engine built as a Hydra delegate meant to progressively replace Mantra, offering native USD integration, a modern architecture optimized for current processors, MaterialX support, and an XPU variant using both CPU and GPU.

In 2019, NVIDIA announced Omniverse, a USD-based collaborative platform letting teams work together on complex 3D projects regardless of the software each uses - integrating a synchronization system for cross-application consistency, a real-time RTX render engine for photoreal visualization, connectors for major 3D software (Maya, 3ds Max, Houdini, Blender), and an extensible microservice architecture. Omniverse represents the convergence of NVIDIA's own technologies (RTX, Iray), open standards (USD, Hydra), and third-party interoperability.

Alongside USD, MaterialX emerged as an open standard for material and shader-graph description - initially developed by Lucasfilm in 2017, then transferred to the Academy Software Foundation - offering an XML/JSON format for shader graphs, independence from specific shading languages, portability across render platforms, and an inheritance/substitution system for reuse and customization. Pixar, NVIDIA and SideFX have all contributed to MaterialX's adoption as USD's standard companion for material exchange.

The Three Companies at a Glance

Notable Collaborations

• Pixar and SideFX - collaboration on integrating USD into Houdini, and joint development of Hydra functionality.
• NVIDIA and SideFX - mutual pipeline support, Houdini optimization for NVIDIA GPUs, and Karma's integration with CUDA and OptiX.
• Pixar and NVIDIA - Omniverse support for Houdini via connectors, USD optimization for NVIDIA GPUs, and support for Storm (Pixar's own Hydra delegate) inside Omniverse.
• Multi-party collaboration - joint work on open standards like MaterialX, the Alliance for OpenUSD (Pixar, NVIDIA, Apple, Adobe, Autodesk and others), the Academy Software Foundation's VFX technology standardization, and multi-renderer support for Open Shading Language.

Timeline Reference Table

Where This Is Heading

By 2025, the VFX rendering ecosystem shows an unprecedented level of interoperability: USD is adopted as a near-universal exchange format across the industry, Hydra allows different render engines to be swapped transparently within one pipeline, MaterialX standardizes material representation across platforms, and platforms like Omniverse enable cross-tool team collaboration. Current and likely future directions include deeper AI integration into rendering workflows (denoising, upscaling, content generation), continued convergence between real-time and offline rendering, hybrid CPU/GPU rendering architectures, and ever more capable open standards.

Each company's likely trajectory: Pixar, having transformed the industry with USD and Hydra, will likely keep opening internal technology - possibly standardizing animation/rigging systems, universal simulation formats, or AI-integration frameworks for creative pipelines. NVIDIA sits at the intersection of hardware acceleration, AI and collaborative platforms, likely continuing to grow Omniverse, deepen AI integration into rendering (content generation, optimization, denoising), and push specialized GPU architectures and distributed compute for rendering. SideFX continues to push procedural innovation and fast integration of emerging technology - likely advances in nodal/procedural tooling, Karma's maturation as a reference Hydra delegate, generative-AI integration into procedural workflows, and next-generation physical simulation.

The most interesting territory may be where these three converge: AI-assisted creation tools built on NVIDIA's AI expertise, Pixar's USD architecture and SideFX's procedural workflows; text- or sketch-driven scene generation and editing that stays USD-compatible; blurring boundaries between CPU/GPU, real-time/offline and local/cloud rendering as RTX, RenderMan and Karma converge toward adaptive, resource-aware rendering systems; universal, physically realistic material libraries built on MaterialX; and Omniverse-style platforms enabling true real-time multi-artist, multi-software collaboration on one scene.

Thirty years of VFX rendering history show an industry moving from fragmented, isolated in-house solutions toward an interconnected ecosystem where interoperability and open standards are the norm. Pixar established the foundations of photoreal rendering and paved the way to standardization with USD; NVIDIA transformed graphics hardware and democratized rendering techniques once reserved for supercomputers; SideFX revolutionized procedural workflows and kept adapting to integrate new technology into its ecosystem. The result: artists can move between tools with less friction, studios can build hybrid pipelines using the best tool for each task, innovation accelerates through shared knowledge and standards, and technology once reserved for major studios is now within reach of independents.