Recently, there has been a steady stream of news about the RTX 40 SUPER series upgrade, and the specs have been revealed.
Now, for the first time, kopite7kimi has given the GPU core number, the number of CUDA cores, and the L2 cache capacity of the three models.
The RTX 4080 SUPER will use a full-blooded version of the AD103, specifically numbered AD103-400, with all 10,240 CUDA cores turned on, an increase of 512 compared to the RTX 4080.
It has been rumored that its video memory will be increased to 20GB GDDR6X, but this means that the memory bit width will be reduced from 256-bit to 160-bit, which is not very reasonable, so it is more reasonable to maintain 16GB GDDR6X.
The RTX 4070 Ti SUPER will indeed have two cores, numbered AD103-275 and AD102-175, with 8448 CUDA cores, an increase of 768 compared to the RTX 4070 Ti.
Plus, it will have 48MB of L2 cache with an unknown memory configuration.
The RTX 4070 SUPER is also a mix of two cores, namely AD103-175 and AD104-350, and the number of CUDA cores has increased by 1280 to 7168, which is quite impressive.
The L2 cache is also 48MB, but the video memory is still unclear.
It is worth mentioning that the RTX 4070 Ti SUPER will be the first model to use both Ti and SUPER, but kopite7kimi said that it has doubts about the name, and it is not ruled out that there will be changes in the future, and even these specifications may change.
