AMD's ML-powered neural rendering technology might not be exclusive to Radeon GPUs, as the company's translation of neural operations will be compatible with other GPU shaders.
AMD's FSR Redstone Won't Be Dependent On AI Acceleration Capabilities, Hence Will Supporter Older GPUs Too
Check out our guide to enable AMD FSR 4 in any game here!
For those unaware, AMD announced the FSR Redstone at Computex 2025, a Machine Learning suite that allows developers to integrate neural rendering technologies into their titles, enhancing visuals and performance. The suite will bring in several ML-focused features, and in a report by 4gamer.net interviewing AMD's Senior Director of Software Development, Chris Hall, it is claimed that FSR Redstone employs AMD's ML2CODE (Machine Learning to Code), which is a part of the ROCm stack, and we'll discuss its importance next.
AMD's ML2CODE has a simple purpose: to take trained neural network models and translate them to GPU compute shader code. It produces optimized HLSL code that can be executed on any GPUs supporting modern shader pipelines, and since Redstone requires runtime ML inference, ML2CODE acts as an intermediary by converting the neural rendering core into standard compute shaders. This means that FSR Redstone shader code can be executed on AMD, NVIDIA, and Intel GPUs, allowing cross-platform support.
FSR Redstone was developed using AMD ML2CODE (Machine Learning to Code), a research project from ROCm. The core part of the neural rendering technology is converted into optimized Compute Shader code by utilizing ML2CODE. This means that FSR Redstone's neural rendering core can also run on GPUs made by other companies.
At AMD, we use HIP in the development process for many innovative new AI-related technologies. ML2CODE aims to integrate with the most commonly used graphics rendering pipelines, such as Vulkan's shader language "GLSL" and DirectX's "HLSL".
It's highly likely that the AI cores of the various AI-related functions used in FSR Redstone are developed using HIP code. This is because HIP code can output code optimized for each generation of Radeon GPU, and thanks to this architecture, it can also run on GPUs other than AMD. Regardless of whether this makes sense, if HIP code is converted to CUDA and built with an NVIDIA compiler, it will likely run on an NVIDIA GPU.
AMD's Senior Director of Software Development, Chris Hall (via 4Gamer)
Interestingly, Hall has also disclosed that AMD's FSR Redstone won't specifically require AI acceleration capabilities at runtime. This means that all ML-focused features will be available to older GPUs since, instead of executing AI cores during runtime, Redstone optimizes the shader code and then executes it, allowing a speed-up without the need for AI compute. Of course, there will be a performance overhead while using Redstone on older hardware, but support is expected to be there.
This is a huge development in rendering technologies, especially for AMD's RDNA tech stack. We have seen that FSR 4 was limited to RDNA 4, leaving behind the older generations. Since Redstone is a first-of-a-kind ML-based implementation by AMD, it could very well support RDNA 3 as well, bringing a performance boost to the platform.
About the author: Muhammad Zuhair is a hardware and technology reporter for Wccftech, specializing in the semiconductor industry and the complex interplay between technology, manufacturing, and geopolitics. His coverage focuses on the corporate strategies and technological roadmaps of industry giants like TSMC, NVIDIA, Samsung, and Intel. Zuhair's expertise lies in deconstructing complex topics such as fabrication nodes (e.g., 2nm process), the economic impact of policies like the CHIPS Act, and the strategic development of AI infrastructure from NVIDIA, AMD and Intel.
Follow Wccftech on Google to get more of our news coverage in your feeds.
