NVIDIA founder and CEO Jensen Huang is scheduled to present a keynote address just before CES 2025 (International CES). On January 6, 2025, at 18:30 Pacific Time, he will introduce the next-generation GeForce RTX 50-series gaming graphics cards. These cards are based on the Blackwell architecture and are produced using a customized TSMC (Taiwan Semiconductor Manufacturing Company) 4NP process. The initial release will feature five products, specifically the RTX 5090, RTX 5080, RTX 5070 Ti, RTX 5070, and a version of the RTX 5090 D tailored for the mainland China market.

This generation of the GeForce RTX 50 series will transition to GDDR7 memory, alongside fully adopting the PCIe 5.0 standard. The series will feature 16Gb (2GB) GDDR7 modules with a speed of 28Gbps. Notably, it has been suggested that the RTX 5080 might include GDDR7 modules operating at 30Gbps, unlike the other models, which would result in a memory bandwidth of 960GB/s with a 256-bit width. This strategy bears some resemblance to the RTX 4080, which utilized GDDR6X at 22.4Gbps, an increase from 21Gbps on other models.

Rumors suggest that AMD's upcoming RDNA 4 architecture and its Radeon RX 8000 series may be more conservative in terms of memory, choosing GDDR6 at a speed of 18 Gbps. Intel's newly launched ARC B580 graphics card also employs GDDR6, albeit at a slightly higher speed of 19 Gbps. From a graphics memory perspective, the GeForce RTX 50-series is likely to be more expensive. Although GDDR7 is available in 24Gb (3GB) modules, NVIDIA appears not to be planning their adoption just yet.