NVIDIA introduced its new Rubin CPX processor at the recent AI Technology Conference recently, with the new CPU being part of the next-gen Vera Rubin NVL144 CPX system... and rumor has it this could be the foundation for its next-gen flagship GeForce RTX 6090 graphics card.
YouTuber "High Yield" is known for die shot analysis, where in a new post on X he noticed that Rubin CPX looks to pack raster units, display engines, and a full set of ROPs... these are components that aren't required for total AI workloads, but are the essential parts of a regular graphics card for rendering. This led him to ponder if Rubin CPX could be the foundation of the next-gen GeForce RTX 6090.
The published die render of Rubin CPX sees it possibly featuring 16 Graphics Processing Clusters that would each have 6 x TPCs (Texture Processing Clusters) for a total of 192 SMs, which is the same amount of SMs as the Blackwell GB202 GPU inside of the RTX 5090. Assuming that the final gaming variant of the GPU has the same configuration as Blackwell (8 x TPCs) that would mean we're looking at 256 Streaming Multiprocessors.
The full and final design of the chip might sport 256 ROPs, a large 512-bit memory bus, and PCIe 6.0 support -- and when compared to the RTX 5090 and its GB202 GPU with 192 SMs and 176 ROPs -- the GR202 should have improvements across the board.
NVIDIA could ship its next-gen GeForce RTX 6090 graphics card with around 28,000 CUDA cores, which is a large 28% increase in CUDA cores over the RTX 5090. If the NVIDIA Rubin CPX ends up being repurposed for consumer graphics cards, the RTX 6090 is shaping up to be a true next-gen gaming GPU beast.



