TECH NEWS – Another update is on the way, and this time Nvidia is launching a Ti Super card as well. It may sound odd, but that’s their call, not ours.
A trusted insider, kopite7kimi, tweeted about two upcoming Nvidia GPUs that have not yet been officially revealed: the GeForce RTX 5070 Ti Super and the RTX 5070 Super. The former will feature the GB203-350-A1 Blackwell GPU on the PG147-SKU55 PCB with 8,960 CUDA cores—unchanged from the non-Super version—but with a 50% VRAM boost, totaling 24 GB instead of 16 GB. Memory bandwidth and bus width will remain the same (896 GB/s using 32 Gbps modules and a 256-bit interface). The RTX 5080 Super will also use 32 Gbps modules, achieving 1 TB/s bandwidth. However, power consumption will increase by 16%, rising to 350W compared to 300W for the standard RTX 5070 Ti. The additional memory will benefit AI tasks and content creation workflows, and the suggested price will be around $750.
GeForce RTX 5070 Ti Super
PG147-SKU55
GB203-350-A1
8960FP32
256-bit GDDR7 24G 28Gbps
350W— kopite7kimi (@kopite7kimi) June 29, 2025
kopite7kimi also leaked details about the RTX 5070 Super, which will use the GB205-400-A1 Blackwell GPU and PG147-SKU65 PCB. This model will see more noticeable upgrades, including a 4.1% increase in CUDA cores (6,400 instead of 6,144). It will also feature 50% more VRAM—18 GB instead of 12 GB—while maintaining the same bandwidth, speed, and bus width (672 GB/s, 28 Gbps, 192-bit). Power consumption will also increase by 10%, from 250W to 275W. This card is expected to launch at a price of around $550.
GeForce RTX 5070 Super
PG147-SKU65
GB205-400-A1
6400FP32
192-bit GDDR7 18G 28Gbps
275W— kopite7kimi (@kopite7kimi) June 29, 2025
Nvidia is likely using 3 GB GDDR7 memory modules on both cards to enable the VRAM increase. These GPUs are expected to be officially announced during CES in early January, though the early leak of specs is undoubtedly an awkward moment for Nvidia.




Leave a Reply