NVIDIA appears to be preparing a second version of the GeForce RTX 5070 graphics card. This potential change in silicon allocation could influence how the manufacturer structures its mid-range product offerings. Consumers may experience variations in performance or availability as the product lineup evolves.
Leaker claims larger die keeps same 6144 CUDA cores and 12GB memory
The rumored variant reportedly uses the GB203 GPU die. Leaker Hongxing2020 stated that the 5070 uses the GB203 core. This contrasts with the current RTX 5070, which relies on the GB205-300 GPU. The GB203 die currently powers the higher-end RTX 5070 Ti and RTX 5080 models.
- GPU: GB203 (variant)
- CUDA cores: 6,144
- Memory: 12GB GDDR7
- Memory bus: 192-bit
- TGP: 250W
Despite the larger die, the technical specifications likely remain unchanged. The new card is expected to keep 6,144 CUDA cores, 12GB of GDDR7 memory, and a 192-bit memory bus. The total graphics power (TGP) is also expected to stay at 250W. There is currently no indication that the GB203 version will have more CUDA cores, more memory or a wider memory bus than the existing RTX 5070.
This strategy mirrors previous releases where NVIDIA used larger dies for lower-tier cards. The RTX 5060 series followed a similar path by using the GB205 die instead of the original GB206. Leaker Hongxing2020 also indicated a new NVIDIA driver release on September 9, potentially linked to this hardware change.
We've been tracking GeForce RTX 5070 closely — see our earlier coverage on GAINWARD GeForce RTX 5070 Snow OC. This report coincides with the broader market availability of Blackwell-based graphics cards as NVIDIA explores different silicon configurations.
The second RTX 5070 variant remains unconfirmed by NVIDIA. The company has not officially released details about the GB203-based model. The September 9 driver date is the only concrete timeline mentioned so far.



Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.