AMD Details Next-Gen Graphics Architecture, Poised for PC and Console Gains
Santa Clara, CA – AMD today outlined key features of its forthcoming graphics architecture, designed to power both next-generation pcs and future game consoles, including Sony’s PlayStation. The new architecture builds upon advancements made with RDNA 4 and introduces several innovations aimed at boosting performance and efficiency.
A core component of the new design will be “neural arrays,” compute units functioning as AI accelerators. While the current PlayStation utilizes an early, experimental version of this technology, the upcoming architecture will substantially expand its capabilities and instruction set support.
AMD is also integrating “radiance Cores,” a dedicated hardware solution for ray tracing acceleration - building on performance gains already achieved with RDNA 4.
Perhaps the most notable growth is “Universal Compression,” a technology designed to compress all data transfers at both the GPU level and between gpus and memory. This contrasts with approaches that compress only select data types. AMD believes Universal Compression contributes to its ability to achieve comparable performance to competitors using faster GDDR7 memory while utilizing the more cost-effective GDDR6, and potentially enabling the use of LPDDR5X in lower-tier GPUs. The architecture’s focus on compression also appears to be influencing cache strategies, with an emphasis on enlarged L2 caches rather than large L3 caches across the entire GPU range.
AMD anticipates the new Radeon GPUs will re-establish the company as a major player in the high-end graphics card market. The architecture’s modularity, allowing GPUs to be used individually or in multi-GPU configurations, is also expected to deliver performance improvements in entry-level APUs.
