GeForce RTX 5070 Ti And 5070 Finalized Specs Allegedly Revealed Ahead Of CES

GeForce RTX graphics cards on a black and green background.

We’re not less than two weeks away from NVIDIA unveiling its next-generation GPUs, with company co-founder and CEO Jensen Huang on the schedule for an opening keynote at CES a day before the event officially commences. The closer we get to a product launch, typically the more reliable the leaks become. With that in mind, we may already know what to expect from at least two of NVIDIA’s upcoming GeForce RTX 50 series SKUs.

In case you missed it, check out our write-up on what to expect from NVIDIA’s keynote. The truncated version is we’re expecting a new generation of desktop graphics cards based on NVIDIA’s Blackwell architecture, with the possibility that we’ll also see some mobile GeForce RTX 50 series GPUs unveiled for laptops.
As it pertains to the desktop, the consensus among leakers is we’ll see NVIDIA unveil its flagship GeForce RTX 5090, along with a GeForce RTX 5080, GeForce RTX 5070 Ti, and GeForce RTX 5070.
Post on X/Twitter outlining specs for the GeForce RTX 5070 Ti.

It’s those 5070 models that are the subject the latest leak, with X/Twitter user kopite7kimi revealing some key details in a series of posts ahead of the event. Starting with the GeForce RTX 5070 Ti, the leaker says it will be based on NVIDIA’s GB203-300-A1 GPU with 8,960 CUDA cores, 16GB of GDDR7 tied to a 256-bit memory bus, and a 300W TBP rating.

Post on X/Twitter outlining specs for the GeForce RTX 5070.

Below that, kopite7kimi indicates that the regular GeForce RTX 5070 model will be built around NVIDIA’s GB205-300-A1 GPU with 6,144 CUDA cores, 12GB of GDDR7 memory linked to a 192-bit memory bus, and a 250W TBP rating.

If those specifications are correct, here’s how they would compare to the current-generation equivalents…

  • GeForce RTX 5070 Ti: 8,960 CUDA cores, 16GB GDDR7, 256-bit bus, 300W
  • GeForce RTX 4070 Ti: 7,680 CUDA cores, 12GB GDDR6X, 192-bit bus, 285W
  • GeForce RTX 5070: 6,144 CUDA cores, 12GB GDDR7, 192-bit bus, 250W
  • GeForce RTX 4070: 5,888 CUDA cores, 12GB GDDR6/6X, 192-bit bus, 200W

Crunching the numbers, we’re looking at a 16.7% increase in CUDA cores for the GeForce RTX 5070 Ti versus the GeForce RTX 4070 Ti, as well as more VRAM and a fatter bus width. The GeForce RTX 5070, meanwhile, is tipped for a smaller 4.3% increase in CUDA cores, while keeping the VRAM and memory bus width the same. Of course, both of the 5070 cards will also benefit from NVIDIA’s Blackwell architecture and GDDR7 memory.

There’s also the GeForce RTX 4070 Super Ti (8,448 CUDA cores, 16GB GDDR6X, 256-bit bus, 285W) and GeForce RTX 4070 Super (7,168 CUDA cores, 12GB GDDR6X, 192-bit bus, 220W), for added context. Those are both more recent additions to the GeForce RTX 40 series, and naturally it’s possible we could see Super variants of the upcoming GeForce RTX 50 series at some point.

Pricing will be key in determining the value proposition (duh!), and it remains to be seen what AMD has up its sleeve for the Radeon RX 9000 series. Meanwhile, Intel has made a strong play in the more affordable segment with its Arc B580 Battlemage GPU, which carries a $249 MSRP. The challenge there, however, is supply enough cards to meet demand.