Select Language:
Sony is preparing to elevate the visual capabilities of the PlayStation 5 Pro by incorporating AMD’s FidelityFX Super Resolution 4 (FSR 4) technology for upscaling. This initiative is part of a partnership between Sony and AMD, referred to as Project Amethyst, which aims to enhance the graphical quality of the console.
FSR 4, which was recently announced alongside AMD’s Radeon RX 9070 and RX 9070 XT graphics cards featuring RDNA 4 architecture, employs machine learning to produce image quality comparable to Nvidia’s Deep Learning Super Sampling (DLSS). Unlike earlier versions, FSR 4 utilizes hardware-based machine learning techniques for upscaling, resulting in improved performance and enhanced visual clarity.
Mark Cerny, the lead system architect for PlayStation, recently discussed the impact of this collaboration with AMD in an interview with Digital Foundry. He highlighted that the neural network and training strategies employed in FSR 4 are significant advancements that surpass the current PlayStation Spectral Super Resolution (PSSR).
The incorporation of FSR 4 into the PS5 Pro is projected for gaming titles set to release in 2026, with ongoing modifications to ensure compatibility with the console’s design. Cerny expressed optimism about the PS5 Pro’s machine learning potential, which boasts a peak capacity of 300 8-bit Tera Operations Per Second (TOPS), aligning it closely with AMD’s latest graphics processing units.
The enhancements in machine learning and ray tracing will serve as pivotal upgrades for the PS5 Pro, potentially deriving from the AMD’s Amethyst project. During the unveiling of RDNA 4, AMD showcased a demonstration that utilized path tracing and neural rendering, along with a denoising capability similar to Nvidia’s ray reconstruction.
Although it remains early to ascertain the overall success of this integration, it represents a significant leap in upscaling technology for the PS5 Pro, promising gamers sharper and more stable visuals in upcoming titles.
ChatGPT
Perplexity
Gemini AI
Grok AI
