The Nvidia GeForce RTX 5070 Ti is expected to be equipped with the GB203-300 GPU, the same processor used in the higher-end RTX 5080 model. This variant of the GB203 GPU is configured for reduced performance to fit within different tiers of the product lineup. The graphics card is anticipated to feature 8,960 CUDA cores and 16GB of GDDR7 memory operating on a 256-bit memory bus. This configuration offers a memory bandwidth of 896 GB/s, representing a substantial increase compared to previous generations. The RTX 5070 Ti is also reported to have a total board power (TBP) of 350W, which is a 23% increase over the RTX 4070 Ti and RTX 4070 Ti SUPER models, both rated at 285W. 

In terms of memory design, the RTX 5070 Ti's 16GB of GDDR7 memory operates at a speed of 28 Gbit/s, achieving a total memory bandwidth of 896 GB/s. This setup marks a 33% improvement in bandwidth over the RTX 4070 Ti SUPER, which initially featured 12GB of GDDR6X memory on a 192-bit bus with a bandwidth of 504 GB/s. The increased memory capacity and speed in the RTX 5070 Ti facilitate faster data transfer rates, essential for high-resolution gaming and professional applications. Additionally, the 256-bit memory bus ensures efficient data handling between the GPU and memory modules, enhancing overall system performance.

Source: Wccftech