The video explains that Nvidia’s open sourcing of PhysX and Flow could enable community-driven improvements and broader compatibility, but hardware limitations, especially on newer GPUs, and the need for significant software adaptation pose challenges. They also discuss potential performance gains through modern APIs like Vulkan and ongoing projects like Zluda, which aim to expand physics support across platforms, highlighting both opportunities and hurdles ahead.
The video discusses Nvidia’s decision to open source their PhysX and Flow GPU source code, highlighting that this move could lead to significant changes in how physics effects are developed and implemented in games. The speakers agree that open sourcing these technologies is a positive step, as it allows other developers and the community to work on compatibility and improvements. However, they also acknowledge that integrating and repairing compatibility issues, especially for older hardware like the RTX 5000 series, will require considerable effort due to hardware limitations and the need for substantial software rewrites.
A key point raised is that Flow, Nvidia’s fluid simulation technology, does not run on CUDA, which complicates cross-platform compatibility. Since Flow isn’t CUDA-based, it could potentially be adapted to other APIs like OpenCL or DirectCompute, but this would involve significant development work. Moreover, Nvidia’s newer GPU architectures, such as the RTX 5000 series, lack the necessary hardware features to support these physics effects directly, as Nvidia intentionally deprecated hardware support for older PhysX features on newer cards. This means that even with open sourcing, hardware limitations will remain a barrier to seamless implementation.
The speakers also discuss the potential for physics effects to become more performant on GPUs through optimization and better utilization. They suggest that current performance issues stem from the reliance on older APIs like DirectX 9, which lack modern compute capabilities and efficient concurrency. Running physics on older APIs often results in context switches and halts in rendering, reducing overall performance. They propose that translating these effects to modern APIs like Vulkan, especially via tools like DXVK, could improve performance by enabling better interoperability and concurrency, leading to more efficient physics simulations.
Furthermore, they consider the implications for game development and modding. With open source physics code, there is potential for physics and flow effects to be integrated into games or mods that previously didn’t support these features. This could extend the lifespan and capabilities of older titles, or inspire new implementations in existing engines. However, actual adoption will depend on the community’s ability to adapt the code and overcome hardware and software limitations, which remains a complex challenge.
Finally, the discussion highlights ongoing efforts like Zluda, a project aimed at bringing CUDA support to other platforms and enabling 32-bit physics support across different systems. The developers behind Zluda are actively working to adapt Nvidia’s physics code for broader use, indicating that innovative solutions are emerging to overcome current barriers. Overall, the open sourcing of PhysX and Flow opens new possibilities, but realizing their full potential will require significant development, adaptation, and hardware considerations.