NVIDIA is reportedly working on its GeForce RTX 5070 Ti graphics card, with new leaks teasing some information on CUDA core counts and power consumption.
The purported GeForce RTX 5070 Ti will rock and roll with a rumored 8960 CUDA cores, 70 Streaming Multiprocessors, and the expected GB203 "Blackwell" GPU. The leaked data suggests the card is called PG147 and SKU 60 internally at NVIDIA, meaning it uses the same board as the RTX 5080 (which is SKU 45) which means we'll see a cut-down GB203 that's used on the RTX 5080.
Leaker "kopite7kimi" has said that the RTX 5070 Ti should feature a 300W TDP, which is a 50W increase over the RTX 5070 and 100W less than the RTX 5080... so right in the middle. We don't know if we're looking at 300W of TGP (maximum) or TDP (thermal design power).
- Read more: RTX 50 for Q1 2025 launch: RTX 5090, 5080 in January, 5070 in February, 5060 in March
- Read more: RTX 5070 rumor: faster than RTX 4070 Ti SUPER in raster, slightly faster in RT
- Read more: NVIDIA GeForce RTX 5070 rumored with an 18GB GDDR7 version (huh?!)
We've already heard and reported leaks on the GeForce RTX 5070 which is expected to debut with 12GB of GDDR7 memory, with leaks teasing 5% more performance in rasterization, and then 5-10% improvements in RT (ray tracing) over the RTX 4070 Ti SUPER.
NVIDIA is also rumored to be working on the GeForce RTX 5070 with 18GB of GDDR7 memory... we can expect to hear more about that in the New Year. For now, we're hearing lots about the mid-range Blackwell gaming GPUs which is exciting, these bad boys will be competing against AMD's last RDNA (4) GPUs before the company shifts into its UDNA architecture moving forward.