Nvidia has taken a significant step in the development community by completely open-sourcing its PhysX and Flow SDKs, including the previously proprietary GPU simulation kernels. This move represents a major shift in Nvidia's approach to its physics simulation technology, potentially breathing new life into older games while opening doors for developers to create custom implementations across various platforms.
The Complete Open-Source Release
Nvidia has now released the complete source code for both PhysX and Flow under the permissive BSD-3 license. While the CPU-side simulation code of PhysX has been open-source since late 2018, the GPU simulation kernels—the core components that enable hardware acceleration—remained proprietary until now. This full release includes over 500 CUDA kernels that support advanced graphics capabilities such as rigid body dynamics, fluid simulation, and deformable objects. Additionally, Nvidia has released the complete shader implementation of Flow, their specialized SDK focused on fluid simulations like fire, gas, and smoke effects.
PhysX and Flow are now open-source under BSD-3 license, including previously proprietary GPU simulation kernels
What PhysX and Flow Actually Do
PhysX is a real-time physics engine that offloads complex calculations to the GPU, taking advantage of its parallel processing capabilities through Nvidia's CUDA platform. The technology has been implemented in numerous games from the 2010s, including Mirror's Edge, Batman: Arkham Asylum, Metro 2033, and Borderlands 2. Flow, meanwhile, is a more specialized SDK designed specifically for fluid simulation mechanics, creating realistic fire, smoke, and gas effects in real-time applications.
PhysX features over 500 CUDA kernels supporting rigid body dynamics, fluid simulation, and deformable objects
Implications for Legacy Games
The timing of this release is particularly significant for users of Nvidia's new Blackwell GPUs, including the RTX 50 series. Nvidia recently discontinued support for 32-bit CUDA on these GPUs, which has caused performance issues with older games that rely on 32-bit PhysX implementations. When these games run on Blackwell GPUs, the physics calculations fall back to the CPU, significantly reducing performance. With the source code now available, modders could potentially develop compatibility layers that would allow these legacy titles to properly utilize PhysX acceleration on newer hardware.
Open-sourcing may enable compatibility solutions for running 32-bit PhysX games on Blackwell GPUs (RTX 50 series)
Broader Potential Applications
With the complete source code now available, developers have unprecedented opportunities to modify and build upon these libraries. While it would be technically possible to decouple PhysX and Flow from CUDA and port them to hardware-agnostic platforms like OpenCL or Vulkan—potentially enabling support for AMD and Intel GPUs—such an undertaking would be extremely complex. Beyond gaming, access to these GPU kernels and shader simulation code could have far-reaching impacts in fields such as graphics engineering, robotics, architecture and design, and animation.
The Future of PhysX in Gaming
Despite this open-source release, it's worth noting that PhysX is largely considered a legacy technology in modern game development. Newer engines, such as Unreal Engine 5 with its Chaos Physics engine, have largely superseded PhysX. The technology is supported in approximately 1,000 games, most of which require Nvidia GPUs to properly render advanced physics effects. Some dedicated gamers have even resorted to installing secondary, low-cost Nvidia GPUs specifically to handle PhysX calculations while using newer RTX 50 cards for primary rendering.
PhysX is supported in approximately 1,000 games, most requiring Nvidia GPUs for hardware acceleration
Community Response and Potential
The developer and modding communities have responded positively to this announcement, seeing it as an opportunity to extend the lifespan of older titles and potentially create more universal implementations of physics simulation technology. With full access to the source code, developers can now study, modify, and optimize these libraries for specific use cases, potentially leading to innovative applications beyond what Nvidia originally envisioned.