GeForce RTX 5070 and RTX 5070 Ti specs reveal cores counts, memory, and more

We're a week away from NVIDIA finally unveiling its upcoming Blackwell-powered GeForce RTX 50 Series at CES 2025. Several credible sources and rumors suggest the company will debut the new line-up with the GeForce RTX 5080 first, followed by the flagship GeForce RTX 5090 a few weeks later. Both GPUs are expected to hit shelves by the end of January 2025, with NVIDIA expected to keep the momentum going with more GPU releases in February.

2

Specifically, the GeForce RTX 5070 and GeForce RTX 5070 Ti. These two cards will greatly interest PC gamers as they target the mid-range or more affordable enthusiast market - NVIDIA's X70 class GeForce GPUs often become one of the best-selling options for each generation. Thanks to known GeForce RTX and NVIDIA insider @kopite7kimi over on X, we now have full details on the hardware configuration of these GPUs.

The GeForce RTX 5070 will feature 12GB of fast GDDR7 memory on a 192-bit bus, matching the capacity of the GeForce RTX 4070. It will also have a CUDA Core count of 6,144, higher than the GeForce RTX 4070's 5,888 CUDA Cores but lower than the RTX 4070 SUPER's 7,168 CUDA Cores.

Of course, being based on NVIDIA's new Blackwell architecture, this isn't a 1:1 comparison. Based on the latest information, we also learned that the RTX 5070 will be a 250W card and the GeForce RTX 5070 Ti a 300W card. The RTX 5070 Ti increases the memory capacity to 16GB of GDDR7 on a 256-bit bus, with 8,960 CUDA Cores - a 45% increase over the RTX 5070.

Based on rumor and insider information, here's a breakdown of the current GeForce RTX 50 Series specs. The key details missing are clock speeds, exact release dates, and pricing.