GFX1156 is landing in Mesa 26.2 and the Linux 7.2 kernel before AMD has named a product, and the patch boundary around Strix Halo tells you where in the lineup it sits.
Before AMD names a product, its drivers are already upstream. That is the operative fact behind this week's Mesa 26.2 patch work plumbing GFX1156, the latest graphics IP block in AMD's RDNA 3.5 family. The work, as Phoronix's Michael Larabel documents, is more than a driver update. It is an engineered disclosure: AMD's engineers are landing user-space and kernel-side support for silicon that has no public product name attached to it.
The new IP block is GFX 11.5.6. Alongside it, a coordinated set of companion blocks is being submitted to the Linux 7.2 merge window: SDMA 6.4, NBIO 7.11.5, IH 6.4, HDP 6.4, MMHUB 3.4.2, SMU 15.0.5, ATHUB 3.4.2, and VPE 2.2. On the user-space side, Mesa's RadeonSI Gallium3D and RADV Vulkan drivers are picking up GFX1156 enablement patches as part of the Mesa 26.2 release cycle, per the Phoronix writeup.
What makes this noteworthy is not the version numbers. It is the boundary the patch draws. The Mesa change narrows AMDGPU_STRIX_HALO_RANGE to end just before GFX1156, which, per Phoronix's reading of the diff, means GFX1156 is not another Strix Point variant. It is a newer APU part that sits after Strix Halo in AMD's IP-block sequence. That is a public, code-visible statement about the APU roadmap, and it is being made through driver plumbing rather than a press release.
The version taxonomy places GFX1156 inside the GFX115x family, the RDNA 3.5 "refresh" line. The same patch series notes that GFX1156 reuses the GFX1153 common code paths. GFX1153 is the block associated with Medusa Point, the APU that has been circulating in leaks as a follow-on to Strix Point. The code-path sharing is the strongest publicly available hint about which APU family GFX1156 sits closest to, but it is a hint, not a confirmation. AMD has not publicly tied GFX1156 to a named commercial product, and that gap is load-bearing. Any reading that collapses the unknown into a confident SKU or codename is doing the reader a disservice.
The pattern is the point. Linux driver development is a leading indicator, not a trailing one, for AMD APU silicon. When GFX1156 surfaces in drm-next, then in mainline Linux 7.2, then in Mesa 26.2, the APU it will eventually power is, in effect, announced through the driver. No keynote, no codename, no product page. Just a commit, a version constant, and a range check. Anyone tracking the kernel mailing list or Phoronix's Mesa coverage can see the next APU arriving months before AMD says a word about it.
For readers, the practical takeaway is a small one. The next time a Phoronix headline names a new GFX block landing in AMDGPU or Mesa, the right question is not "what is the patch doing" but "what product is the patch pointing at." The answer is usually in the version number, the IP-block neighbors, and the ranges the diff trims. GFX1156 is the cleanest recent example: it tells you a new RDNA 3.5 APU is on the way, and that AMD's open-source-first workflow will name it on its own schedule, in the diff stat, long before the marketing team gets involved.
What to watch next is the Linux 7.2 merge window itself. If GFX1156 and its companion blocks land cleanly in the kernel, and the Mesa 26.2 user-space patches follow, the unnamed APU behind them is, in any meaningful sense, already public. The announcement will just be the part that has not happened yet.