‘Could this be the RTX 6090’: Tech analyst reckons Nvidia’s just-announced AI GPU hides a tantalizing secret that could throw down a massive gauntlet to AMD
- Rubin CPX announced with 128GB GDDR7 memory and advanced PCIe 6.0 support
- Speculation builds that Rubin CPX architecture could underpin Nvidia’s next GeForce flagship
- Nvidia insists Rubin CPX is for AI but questions remain about gaming potential
Nvidia recently introduced its Rubin CPX GPU, an AI compute processor built for long-context workloads.
It was shown as part of the Vera Rubin NVL144 CPX system and features 128GB of GDDR7 memory along with PCIe 6.0 support.
The company presented it as an engine for inference in fields such as research, development, and high-definition media generation.
Nvidia RTX 6090?
Despite its AI positioning, the design has sparked speculation about whether Rubin CPX could also form the basis of a future GeForce graphics card.
In reporting about the new GPU, we joked about Nvidia selling us one to run games, but we weren’t the only people who saw its gaming potential.
On X, YouTuber High Yield remarked, “Some quick assumptions about Nvidia’s new Rubin CPX. Quite similar to GB202, with some obvious changes to the GPCs. Looks like it has full raster units (full GPU) with up to 256 ROPs. Could this be the RTX 6090? (even though Rubin was supposed to be AI only).”
VideoCardz was intrigued about High Yield’s suggestion, agreeing that Rubin CPX, shown during the AI Technology Conference, appears to include elements not typically needed in AI compute processors.
These include raster units, display engines, and a complete set of ROPs, all of which are essential in gaming GPUs.
The site noted that Rubin CPX could integrate 16 Graphics Processing Clusters, each with 6 Texture Processing Clusters, totaling 192 SMs.
A gaming version following the Blackwell pattern of 8 TPCs per GPC could expand this to 256 SMs.
VideoCardz also pointed out that such a configuration might support 256 ROPs, a 512-bit GDDR7 interface, and close to 28,000 CUDA cores.
Compared with the GeForce RTX 5090’s GB202 GPU, with 192 SMs and 176 ROPs, this would offer around a 28 percent increase in cores. That scale would place any resulting RTX 6090 among the largest consumer GPUs ever built.
Rubin CPX was shown purely as an artistic render rather than a confirmed die shot, which is hardly surprising given that it’s not expected to arrive until late next year.
Nvidia has described Rubin CPX strictly as an AI-focused product, and if there is any hidden GeForce connection, it will be quite a wait before we know for certain.
Some quick assumptions about @Nvidia’s new Rubin CPX. Quiet similar to GB202, with some obvious changes to the GPCs. Looks like it has full raster units (full GPU) with up to 256 ROPs. Could this be the RTX 6090? (even though Rubin was supposed to be AI only) pic.twitter.com/EBRYpH39VBSeptember 10, 2025
You might also like
Rubin CPX announced with 128GB GDDR7 memory and advanced PCIe 6.0 support Speculation builds that Rubin CPX architecture could underpin Nvidia’s next GeForce flagship Nvidia insists Rubin CPX is for AI but questions remain about gaming potential Nvidia recently introduced its Rubin CPX GPU, an AI compute processor built for…
