AMD’s next-gen Zen 6-based ‘Medusa’ CPU rumors: 2025 teased, some packaging tech from MI450 GPU

AMD is about to launch its next-generation Zen 5 CPU architecture inside of Ryzen AI 300 series “Strix Point” APUs and Ryzen 9000 series “Granite Ridge” desktop CPUs… while we’ve got some juicy new details on the future-gen Zen 6 “Medusa” CPUs.

In a new video from leaker Moore’s Law is Dead, we’re hearing new rumors about Zen 6 “Medusa” CPUs and it looks like it’s possible they’ll tape out by Q2 2025, which isn’t far away at all now, with production to follow “before the end of the year” according to MLID’s first source. This source added that if things go well, production would begin by the end of 2025, but “it could slip” into 2026.

This source also teased that AMD’s next-gen Zen 6 “Medusa” products could use the same packaging technology as AMD’s next-gen Instinct MI450 AI accelerator, which we have been hearing rumors about lately. Another source told MLID that “Medusa” is a Zen 6 product that will launch on desktop AM5, and that it could be a family that launches on laptops as well.

Source 1: “I (AMD) just heard that“Medusa”will tapeout by Q2 of 2025, and production should follow before the end of the year if things go well (it could slip to 2026). I was also told that this is a Zen 6 product that uses some of the same packaging tech as MI450“.

Source 2: “As far as I (works with AMD) am aware,“Medusa”is a Zen 6 product that will launch on desktop AM5. Also, I believe it may be a family that launches to laptop as well. Oh, and from what I hear both Strix and Medusa Halo now use N3E IODs, and may fit AM5“.

Source 3: “I’ve (AMD partner) heard of“Medusa”. In fact, it sounds like there’s“Medusa Ridge”,“Medusa Halo”, and“Medusa Point”. NOT 100% sure, but those would all be“client”.

Source 4: “I (AMD) don’t remember hearing about“Medusa”, but I rarely hear those sorts of codenames. However with regards to a 3nm IOD for Strix/Medusa – I will say that I DO know that Zen 5 Strix was originally planned to be a 3nm chiplet architecture, but that got axed after we had tons of issues, but also after TSMC 3nm got delayed! And now that I think about it… STX-Point will launch 2 quarters late, and yet STX-Halo is poised to launch a year late. That’s a bigger delay, it can’t just be from chiplet issues“.