AMD’s Next-Gen GPUs Will Be RDNA 5, Not UDNA

According to Wccftech, AMD has added initial support for a GPU architecture identified as GFX13 to the LLVM compiler stack, explicitly linking it to the “RDNA 5” architecture. This confirms the next-gen architecture will continue the RDNA naming convention instead of switching to “UDNA.” The GFX1310 branch within the code points to the discrete GPU lineup, which would logically become the Radeon 10000 series. This early work in LLVM is critical for future driver and toolchain optimizations. The RDNA 5 GPUs are not expected to launch before 2027 and are slated to utilize TSMC’s N3P manufacturing process. This timeline would put them in direct competition with NVIDIA’s anticipated RTX 60 series “Rubin” GPUs around mid-2027.

RDNA 5 Confirmed

So, the naming debate is settled. For anyone following the leaks, seeing “GFX13” tied directly to “RDNA 5” in the official LLVM documentation is about as clear a confirmation as you can get before an announcement. This isn’t some random entry in a database; LLVM is the backbone of AMD’s open-source software stack for Linux and development tools. Adding an architecture here means the foundational engineering work has started. It’s a long lead-time item. Basically, they’re laying the plumbing now so the software house doesn’t flood when the hardware arrives years later.

The 2027 Showdown

Here’s the thing: a 2027 launch window tells us a lot. It suggests AMD is planning a major architectural leap, not just a refresh. Jumping to TSMC’s N3P node is a big deal, and it lines up almost perfectly with NVIDIA’s expected Rubin timeline. We’re looking at a potential head-to-head battle in a new process generation. After the reported scaling back of the high-end for RDNA 4, all eyes will be on whether RDNA 5 marks AMD’s return to fighting for the absolute performance crown. Can they finally mount a sustained challenge? The code name is a promise, but the execution is everything.

What It Means For The Market

This early visibility is a double-edged sword. On one hand, it shows AMD is being proactive with its software ecosystem, which has been a pain point. On the other, it locks in a very long public roadmap. NVIDIA operates with more secrecy, so AMD is constantly having its future plans dissected years in advance. The real impact, though, will be on the competitive dynamics. If RDNA 5 is a genuine high-end contender in 2027, it could finally apply the pricing pressure in the premium segment that gamers have wanted for years. But that’s a huge “if.” For now, it’s a line of code in a compiler. A very telling one.