Specifications for the NVIDIA GeForce RTX 4070 and RTX 4070 Ti have leaked.
In fact, NVIDIA will release not one, but two new GeForce RTX 40-series “Ada” SKUs in January. One of these is the RTX 4070 Ti, which was widely known to be the RTX 4080 12 GB rebranded in response to criticism that compelled NVIDIA to “unlaunch” it. As it turns out, the other has an intriguing set of specifications and is called the RTX 4070. The new RTX 4070 is considerably smaller than the RTX 4070 Ti and is built on the same 4 nm AD104 silicon. When compared to the 7,680 CUDA cores that the fully-loaded RTX 4070 Ti boasts, NVIDIA only enabled 46 of the 60 streaming multiprocessors (SM) that are physically present on the silicon, yielding 5,888 CUDA cores. This is the same number as the RTX 3070 from the previous generation.
In addition to 5,888 CUDA cores, the GeForce RTX 4070 has 46 RT cores, 184 Tensor cores, 184 TMUs, and 64 ROPs, down from 80 on the RTX 4070 Ti. The memory setup is the same as the RTX 4070 Ti, consisting of 12 GB of 21 Gbps GDDR6X memory spread throughout the chip’s 192-bit memory interface for a total memory bandwidth of 504 GB/s. This SKU’s board power, which is rated at 250 W as opposed to the RTX 4070 Ti’s 285 W and the RTX 3070’s 220 W from the 8 nm generation, is an intriguing feature.
