According to reliable information, the upcoming NVIDIA GeForce RTX 40 Super series, built on the Ada Lovelace architecture, is expected to offer enhanced specifications compared to the current GeForce RTX 40 series. However, it appears that there won't be significant changes in the memory configuration. Notably, there's no indication of a 20GB version of the GeForce RTX 4080.
Earlier reports have hinted at the inclusion of three models in the GeForce RTX 40 Super series: the GeForce RTX 4080, RTX 4070 Ti, and RTX 4070. However, the precise specifications of these graphics cards remain unconfirmed.
For the GeForce RTX 4080 Super, which employs the AD103-300 chip, it is expected to feature 10,240 CUDA Cores, an increase from the original 9,728 CUDA Cores. Furthermore, the presence of the GeForce RTX 4080 SUPER is indicated in the "Version History" page of HWiNFO.
Moving on to the GeForce RTX 4070 Ti Super, it shares the same chipset as the GeForce RTX 4080 Super, using the codename AD103-275. This model is projected to house 8,448 CUDA Cores, up from the original 7,680 CUDA Cores. However, there is a mention on Kopite7kimi's Twitter that the GeForce RTX 4070 Ti Super may potentially utilize the AD102-175 chip, which introduces some ambiguity surrounding its chipset designation.
Lastly, the GeForce RTX 4070 Super is expected to feature either the AD104-350 or AD103-175 chip. This graphics card is anticipated to come equipped with a configuration of 7,168 CUDA Cores, marking a significant increase from the original 5,888 CUDA Cores.
Kopite7kimi's track record in providing accurate information is noteworthy, and it aligns with information from one of our sources. However, the final details and specifications will be confirmed once NVIDIA releases the official documentation.
