AMD’s RDNA 4 Architecture: GFX1200 & GFX1201 GPUs Unveiled, Potentially Navi 44 & Navi 48 for Radeon RX 8000 Series

The latest LLVM-Project notes have revealed mentions of two AMD RDNA 4 “Navi 4X” GPUs, specifically the GFX1200 and GFX1201, in the context of defining target names and ELF numbers for new GFX12 targets. These targets include the GFX1200 and GFX1201, with both GPUs currently behaving similarly to the GFX11 “RDNA 3” family. However, significant differences are anticipated in the future, given that the launch of these GPUs is still a year away.

More About AMD RDNA 4 Architecture

Previously, the GFX1200 ID was associated with the Navi 41 SKU, but reports suggest that AMD has canceled its high-end Navi 4X offerings, initially designed to utilize the Navi 4C die for a chiplet-style design. Instead, the company is now reportedly focusing on the high-end and mainstream segments. Notably, AMD traditionally considers products priced above $500 as enthusiast-class, typically utilizing the top Navi x1 dies, but this may change with the upcoming GPUs.

According to a post by @Olrak29_, who has previously provided die block diagrams and configurations for RDNA GPUs, it is speculated that the two RDNA 4 “Navi 4X” IDs, namely Navi 44 and Navi 48 SKUs, could be associated with the GFX1200 and GFX1201. While the nomenclature might suggest low-end SKUs, AMD has clarified that numbering does not necessarily represent performance or segment positioning, and further information is needed before making speculative conclusions.

This development marks the initial stages of work on AMD’s next-gen RDNA 4 architecture, expected to power the Radeon RX 8000 series. The company aims to enhance performance compared to the RDNA 3 family, introducing new features for the Navi 4X class of chips, such as a revamped graphics pipeline, improved ray tracing capabilities, enhanced efficiency, and smarter utilization of AI accelerators. The next-generation AMD Radeon GPU lineup based on the RDNA 4 architecture is anticipated to launch by 2024, utilizing a new and advanced process node.