Multiple recent reports indicate AMD will continue to use RDNA 3.5 for mainstream integrated GPUs (APUs) through 2029 while reserving RDNA 5 for a "premium" class of iGPUs in higher-end SoCs. Industry reporting and insider posts describe a roadmap that segments entry/mainstream APUs on RDNA 3.5 and moves premium/halo APU SKUs to RDNA 5, with Medusa-family parts cited as a transition point. [1][3] Separately, low-level tooling and repository changes show early support for AMD's next-generation RDNA 5 family: patches and identifiers for GFX13 (RDNA 5) and a gfx1310 dGPU ID have been observed in LLVM and noted by hardware news sites and developer analysis, which suggests AMD (or contributors working on AMD support) are already integrating RDNA 5 into compilers and drivers even if products are not yet public. [5][4] At the same time, reporting on market timing suggests AMD may be hesitant to push new discrete GPU launches into a difficult memory and pricing environment — one report claims AMD is pausing new GPU launches until 2027, a decision framed as a reaction to market conditions rather than a technical lack of readiness. [2] Taken together, the coverage points to a two-track situation: AMD appears to be developing and enabling RDNA 5 in software and toolchains while keeping RDNA 3.5 as the workhorse for mainstream APUs through 2029, and simultaneously weighing market timing for discrete-GPU launches. For consumers and OEMs this implies incremental iGPU continuity in mass-market devices, potential premium iGPU upgrades on select SoCs, and uncertainty about the calendar for new discrete Radeon launches driven by supply and strategic timing considerations. [1][3][5][2]