Zluda run unmodified CUDA on non Nvidia hw (www.phoronix.com)

Zluda has announced the release of ZLUDA 6-preview.48, which introduces compatibility with NVIDIA's CUDA 13.1, enabling CUDA applications to run on non-NVIDIA GPUs. This multi-GPU implementation marks a significant step forward, as it aims to support a wider range of hardware without requiring modifications to existing CUDA software. The project particularly emphasizes enhancements for AI workloads, which can benefit immensely from broader hardware accessibility. The significance of this development lies in its potential to democratize CUDA's functionality, allowing developers and researchers to utilize a variety of GPU infrastructures for AI model training and inference. By running unmodified CUDA applications on alternative GPU hardware, Zluda is poised to facilitate innovation in the AI/ML community, making powerful computational resources more accessible. As the project continues to evolve, particularly looking toward 2026, it could reshape the landscape of AI computing by enabling greater flexibility in hardware choice and optimizing performance across diverse systems.