The Gaming Brief:
- AMD engineer Frank Min added a “GDDR7” identifier to the AMDGPU Linux driver, the first direct GDDR7 reference in AMD’s graphics software stack
- The patch isn’t tied to a specific GPU yet, but arrives alongside new IH 8.0 and NBIF 7.10 hardware-block support widely associated with RDNA 5
- Separately, Weibo leaker Golden Pig Upgrade says 4GB and 6GB GDDR7 memory chips are planned for 2027–2028, up from today’s standard 2GB modules
- Denser chips could let narrow 128-bit memory buses carry far more VRAM without redesigning the GPU’s memory interface
AMD has started preparing its open-source AMDGPU Linux driver to recognize GDDR7 memory, according to VideoCardz, which cited a patch submitted by AMD engineer Frank Min and reviewed by Hawking Zhang.
The change adds “GDDR7” to the driver’s list of supported VRAM types, sitting alongside GDDR6, HBM3E, and HBM4 entries.
Current Radeon RX 9000 “RDNA 4” cards use GDDR6, while Nvidia’s GeForce RTX 50 series already runs GDDR7, making this patch the clearest signal yet that Team Red is moving to match its rival.
Arriving Alongside RDNA 5-Linked Hardware Blocks
As VideoCardz details, AMD hasn’t linked the GDDR7 update to a specific graphics card line yet.
However, the patch landed alongside support for IH 8.0, a new Interrupt Handler revision, and NBIF 7.10, a new Bus Interface version, both submitted by AMDGPU maintainer Alex Deucher.
Phoronix, which first reported the patches, links this hardware-enablement work to RDNA 5, though it stresses this is driver-development activity rather than an official AMD product announcement.
A memory-type identifier alone doesn’t reveal the VRAM capacity, bus width, or clock speed for any upcoming card.
Bigger VRAM Pools Without Wider Buses
Separately, Notebookcheck reports that leaker Golden Pig Upgrade expects 4GB and 6GB GDDR7 chips to become available in 2027 or 2028, moving beyond the 2GB modules most current GDDR7 cards use.
The jump matters most for mid-range and budget GPUs built on narrow 128-bit memory interfaces, where VRAM has typically been capped low.
A 128-bit card using four 2GB chips today tops out at 8GB; the same layout with 4GB chips would reach 16GB, and 6GB chips would push it to 24GB, without changing the underlying bus at all.
Existing 3GB Chips Are Already Blamed for Delaying the RTX 50 Super Refresh
VideoCardz reports that current 3GB GDDR7 chips remain expensive, a cost that has reportedly contributed to delays around the rumored RTX 5050 9GB and RTX 50 Super refresh.
At the same time, street prices for the flagship RTX 5090 have hit multi-thousand-dollar heights.
That pricing pressure is the reason why reports suggest 4GB chips could become the more practical mainstream option, while 6GB packages may first appear on workstation and high-memory GPUs instead.
No specific graphics card, from either AMD or Nvidia, has been confirmed to use the new chip densities, and both reports caution the timeline depends on the memory market stabilizing.
