In a bold strategic shift, AMD is planning to unify its datacenter and consumer graphics architectures, merging its CDNA and RDNA frameworks into what it refers to as “New DNA.” This initiative aims to create a unified architecture that will streamline development for both AMD’s enterprise and consumer GPU offerings. According to Jack Huynh, AMD’s senior vice president, this consolidation will eliminate the complexity currently faced by developers who must navigate two distinct architectures. The goal is to attract a wider range of developers by simplifying the development process and focusing on mainstream products rather than just flagship GPUs.
Huynh likened AMD’s effort to Steve Ballmer’s well-known push for developer support, suggesting that a unified architecture will enhance AMD’s appeal to software creators. He acknowledged past challenges with RDNA development, including issues related to memory hierarchy and optimization, but expressed confidence that the new approach will address these problems. By avoiding frequent resets in optimization and focusing on advanced planning, AMD aims to provide a more stable and efficient development environment.
The new architecture is also expected to address AMD’s current limitations in AI capabilities. While Nvidia’s GPUs are renowned for their strong AI support through CUDA and DLSS, AMD’s revisions may bring significant improvements in this area. As AMD prepares to roll out these changes, more details will likely emerge about how they plan to challenge Nvidia’s dominance in AI-driven applications and other high-performance tasks.