NVIDIA is preparing its next-generation GeForce RTX 60-series graphics cards powered by the Rubin architecture, continuing its two-year product cadence. According to reports, the lineup will stick with TSMC's 3 nm FinFET process rather than jumping to sub-2 nm nodes. The flagship RTX 6090 will use the GR202 silicon with 192 streaming multiprocessors and maintain its 512-bit GDDR7 memory interface.

Performance targets include modest raster gains of 30-35% but a massive 100% improvement in real-time path tracing thanks to 5th generation RT cores. The new 6th gen Tensor cores aim to bring DLSS 5 neural rendering to mainstream cards, not just flagship models. This addresses NVIDIA's earlier DLSS 5 demo that required dual RTX 5090 cards.

The mid-range RTX 6080 will feature the GR203 silicon with a 320-bit memory bus and 20 GB of GDDR7, delivering at least 25% more bandwidth. The RTX 6070 gets a significant upgrade with the GR205 chip, 16 GB of memory, and 33% more bandwidth. Memory configurations and bus widths will vary across the lineup to balance performance and cost.