Solo developer wires ZLUDA to AMD's HIP, getting multiple CUDA libraries running on a Radeon RX 9060 XT in Windows

A solo developer has successfully bridged ZLUDA to AMD's HIP SDK, enabling multiple CUDA libraries to run on an AMD Radeon RX 9060 XT in Windows. This breakthrough leverages AMD's latest ROCm updates,... Read original →

This development underscores AMD's strategic expansion in the consumer GPU ecosystem. By bridging ZLUDA with HIP, developers can now run AI frameworks like PyTorch without CUDA, signaling AMD's growing independence from NVIDIA's dominance. Though cuDNN remains unsupported, successful training of a 2.2M-parameter model shows ROCm's rapid software evolution. From a supply chain perspective, this reduces enterprise reliance on NVIDIA, especially amid geopolitical tensions, lowering risks of tech blockades. NVIDIA may intensify its closed CUDA ecosystem, while AMD could leverage ROCm as a mainstream AI development platform. Over the next 12 months, continued community contributions may yield a stable cross-platform toolchain, reshaping GPU market dynamics.