Rubin CPX Sparks Talk of the Future RTX 6090

At the AI Technology Conference, NVIDIA revealed a new processor named Rubin CPX. Officially, it is designed for artificial intelligence. Yet its architecture tells another story. Inside the chip, analysts have spotted components that seem more suited for gaming graphics than machine learning. Raster units, display processors, and a full set of ROPs are present. None of these are necessary for AI tasks but they are essential for rendering 3D worlds. Hardware analyst and YouTuber High Yield believes this is no coincidence. He argues Rubin CPX could form the backbone of the next GeForce flagship, the RTX 6090.

A Peek at Possible Performance

The published render of Rubin CPX shows 16 graphics processing clusters, each housing 6 texture processing clusters. That equals 192 streaming multiprocessors, the same number as the GB202 chip in today’s RTX 5090. If NVIDIA expands each cluster to 8 TPCs, similar to the Blackwell generation, the total would jump to 256 SMs. This would bring a matching 256 ROPs, a 512-bit GDDR7 memory interface, and PCIe 6.0 support. Compared with the 176 ROPs of the GB202, the improvement looks substantial. Even if NVIDIA disables around 10 percent of the cores for yield reasons, the RTX 6090 could still reach about 28,000 CUDA cores. That is nearly 30 percent more than its predecessor.

Waiting for the Next Leap

For now Rubin CPX exists only as a stylized render. NVIDIA plans to launch it toward the end of 2026. Whether it stays an AI processor or evolves into the heart of a consumer graphics card, it already fuels speculation about a powerful new era of GeForce.