GeForce RTX 4070 Ti and RTX 4070 Specifications Leaked – Nvidia

Now, it’s coming Generally reliable leaker Harukaze5719 with a Twitter post to reveal the characteristics of the “younger” GeForce RTX 4070, and it’s disappointing that it has the same number of CUDA cores – if its information is indeed correct – as the outgoing GeForce RTX 3070. It has a smaller bandwidth memory, 192 bits.

Despite the smaller memory bus width, the larger and faster memory, plus the new “Ada Lovelace” architecture results in increased performance, 50% more than TFLOPs, while the memory bandwidth is also higher, approaching 10%.

The new GeForce RTX 4070 GPU is essentially a “cut” version of the AD104 GPU that also “powers” the “Ti” version, has a die that occupies a surface area of ​​292 mm² and is manufactured using TSMC’s N4 method. Although it is much smaller than its predecessor, it has the same number of CUDA and Tensor cores -5,888 and 184 respectively. The new GeForce RTX 4070 GPU is also said to have the same boost clock as the “Ti” version, at 2.6GHz, but due to its “cut-in characteristics”, it has a reduced consumption of 35W with a TDP of 250W.

according to Harukaze5719The new GeForce RTX 4070 delivers 125 GFLOPs per watt, which is slightly better than the GeForce RTX 3070’s 92 GFLOPs per watt. Its bigger brother, the GeForce RTX 4070 Ti delivers 140 GFLOPs per watt. Anyway, apart from the confirmed color specs of the GeForce RTX 4070 Ti which basically shows it to be an RTX 4080 12GB, the specs of the GeForce RTX 4070 are not confirmed and so we have to be careful.