HN
Today

Zluda 6 release (run unmodified CUDA applications on non-Nvidia GPUs)

Zluda 6 arrives, empowering users to run unmodified CUDA applications on non-Nvidia GPUs, effectively bypassing vendor lock-in. This latest update significantly improves Windows support, enhances machine learning capabilities, and surprisingly adds pre-alpha PhysX and Blender texture support. The project, now a passion-driven endeavor after commercial funding ceased, continues to chip away at Nvidia's ecosystem dominance, resonating deeply with HN's technical and open-source-minded audience.

15
Score
0
Comments
#6
Highest Rank
5h
on Front Page
First Seen
Jun 30, 1:00 PM
Last Seen
Jun 30, 5:00 PM
Rank Over Time
11861015

The Lowdown

The Zluda project, which enables CUDA applications to run seamlessly on non-NVIDIA GPUs, has released its latest major update, Version 6, covering developments from Q1 and Q2 2026. This update marks significant progress across several fronts, moving Zluda closer to being a comprehensive alternative for GPU computing.

  • Version 6 Release: While Zluda follows continuous development, Version 6 signifies substantial progress, encouraging users to revisit the project.
  • PhysX Pre-alpha Support: A notable addition is the pre-alpha support for 32-bit PhysX, allowing older games that rely on this technology to run with improved frame rates and visual effects on AMD GPUs.
  • Blender Texture Support: Zluda now includes basic texture support, which is sufficient to enable Blender to function on Zluda, expanding its utility for 3D content creation.
  • Improved Windows Experience: Windows support has seen considerable enhancements, with better handling of performance libraries and clearer guidance for users, addressing a historical gap compared to its Linux counterpart.
  • Enhanced ML Capabilities: Continuous improvements have been made to machine learning support, driven by user feedback and traces from PyTorch users, including numerous compiler bugfixes and performance library optimizations.
  • Project's New Direction: The project has returned to being a weekend passion project after commercial funding ceased. This shift has altered development priorities, focusing on features the developer finds personally 'entertaining', leading to the inclusion of PhysX, Blender, and improved Windows support that were previously not on the roadmap.

Despite a potential shift to less frequent updates due to its passion-project status, Zluda continues to evolve, offering a compelling open-source solution that challenges Nvidia's closed CUDA ecosystem by bringing hardware flexibility to compute-intensive applications.