Please note that this post is tagged as a rumor .
VEGA RDNA3.5 lives
AMD’s integrated graphics roadmap has become hard to follow. The current APU lineup keeps circling back to the same iGPU architecture, while what comes next is a big unknown, or maybe it isn’t?
Older rumors suggested AMD had bigger RDNA 4 plans than what shipped. RDNA 4 was linked to a wider Radeon RX 9000 stack, including a higher-end design often referred to as Navi 4C. That never happened. AMD ended up launching two RDNA 4 GPUs, Navi 48 and Navi 44, which now power a short list of Radeon RX 9000 and Radeon PRO products.
On the integrated side, the story was simpler from the start. RDNA 4 was not expected to land in APUs. In fact, there were no plans to bring RDNA4 as mobile dGPU platform, which is why there has been no Radeon RX 9000 Mobile lineup to pair with Strix Point systems.
For 2026, AMD seems to be focusing on refreshes. Ryzen AI 300 “Strix Point” became Ryzen AI 400 “Gorgon Point.” Ryzen AI MAX 300 “Strix Halo” is expected to become “Gorgon Halo.” All of these are still tied to RDNA 3.5 iGPUs.
AMD is dividing their APU roadmap into:
a) Products targeting lower end markets or markets that don't need good iGPU performance (i.e office laptops, and also laptops using high-end dGPUs). These will continue to use RDNA 3.5 until 2029.
b) "Premium" iGPU products with RDNA5. https://t.co/dfPw4G2Ee9
— Kepler (@Kepler_L2) January 24, 2026
Since yesterday there has been a debate about AMD RDNA strategy. After known leaker Golden Pig Upgrade shared an AI-generated roadmap, some decided to add some context. The slide shows an “iGPU architecture roadmap” where AMD stays on RDNA 3.5 from 2024 through 2028+.
Kepler_L2, known for accurate ADM leaks, added that AMD is dividing its APU roadmap into two tracks: products aimed at lower-end markets, or designs that do not need a strong iGPU, would stay on RDNA 3.5 until 2029. Separate “premium” iGPU products would move to RDNA 5.
While Kepler does not mention what “premium” here means, some rumors suggest that the next-gen Medusa platform will split between Medusa Point, still using RDNA3.5, while the new “Halo/Premium” would move to RDNA5. Mainly due to the fact that these premium chips are said to be using separate GPU tiles.
Still, compared to what is happening at Intel, it looks like AMD has either forgotten to innovate, or at least forgotten how to meet an annual update cadence. Eventually Intel will split its product stack to its own Xe architecture and NVIDIA GPU tiles, as announced last year.
Source: Kepler_L2
