AMD Working on New AI-Based Upscaling Technology

Mark Papermaster, the CTO of AMD, has confirmed that the company is developing a new upscaling technology that utilizes artificial intelligence (AI). One key difference between AMD FSR and its competitors, NVIDIA DLSS and Intel XeSS, is AMD's cautious approach to incorporating AI into the upscaler's pipeline. While DLSS and XeSS use AI deep neural networks (DNNs) to address temporal artifacts in their upscalers, AMD has refrained from doing so. The AMD Radeon RX 7000 series GPUs and Ryzen 7000 CPUs are the first to feature accelerators or ISA that accelerate AI workloads. With a significant user base already using the RX 7000 series, AMD is now looking towards AI for the next iteration of its FSR upscaling technology. Papermaster discussed the company's plans for AI in upscaling technologies in an interview with No Priors.

When asked about exploring AI for upscaling, Papermaster stated: "2024 is a significant year for us as we have invested many years in enhancing our hardware and software capabilities for AI. We have recently enabled AI across our entire product portfolio, including cloud, edge, PCs, embedded devices, and gaming devices. We are now enabling gaming devices to utilize AI for upscaling, and 2024 will be a major year for deployment." Papermaster outlined AMD's two-step approach to integrating AI, starting with a focus on hardware. In 2022-23, AMD introduced ISA-level AI support for Ryzen 7000 desktop processors and EPYC "Genoa" server processors. For laptops, it launched the Ryzen 7040 and 8040 series mobile processors with NPUs for accelerated AI support, along with AI accelerators in the Radeon RX 7000 series RDNA 3 GPUs. Additionally, AMD introduced the Ryzen AI stack for Windows PC applications to enhance client productivity experiences using AI. The company is set to implement AI across its technologies in 2024, with Papermaster confirming the development of a new-generation FSR that incorporates AI.