AMD RDNA architecture page (generation comparison table and RDNA 4 overview)

Vendor architecture overview page. No publication date printed, so published is null. Used for three things the per-SKU pages omit. (1) The generation comparison table, verbatim: columns RDNA / RDNA 2 / RDNA 3 / RDNA 4 with row 'Process | 7nm | 6~7nm | 5~6nm | 4nm', row 'Compute Units | 1st Gen | 2nd Gen | 3rd Gen | 4th Gen', row 'Raytracing | | 1st Gen | 2nd Gen | 3rd Gen', row 'AMD Infinity Cache | | 1st Gen | 2nd Gen | 3rd Gen', and an AI Engine row showing 1st Gen and 2nd Gen only. (2) Explicit binding of RDNA 4 to the RX 9000 family: 'With the latest AMD RDNA 4 architecture, experience next-generation advancements in performance and efficiency with faster raytracing and AI acceleration featured in the AMD Radeon RX 9000 Series graphics.' (3) The AI-accelerator generation: 'Supercharged with AI - 2nd Generation AI' and 'AMD RDNA 4 Architecture ... pushes new boundaries with up to 64 advanced AMD RDNA 4 compute units, featuring a huge generational leap in performance and efficiency-per-CU, 3rd generation raytracing accelerators ...'. Footnote 1 reads 'Based on specifications of AMD RDNA 4 architecture compared to AMD RDNA 3 architecture as of December 2024. RX-1143'. IMPORTANT: AMD's own comparison table gives RDNA 3 as a process RANGE ('5~6nm'), not a single node, which is why process_nm is left null for the RX 7900 XTX rather than filled with a guess. Accessed 2026-10-03. KIND NOTE: this is AMD's own architecture marketing page, filed as 'blog' because the schema's kind enum has no 'marketing' member. The process-node row in its generation table is the weakest-sourced datum on this record and is the reason the referencing accelerator leaves process_nm null.