NVIDIA’s budget Lovelace offerings are turning out to be a bust. The GeForce RTX 4070 Ti is a solid generation upgrade but costs $200 more than the 3070 Ti. The RTX 4070 will feature the same shader count as the RTX 3070, while the RTX 4060 Ti will have a lower shader strength than its predecessor.
The RTX 4070 will leverage the AD104 die, packing 5,888 FP32 cores alongside 12GB GDDR6X memory. The core will be clocked at around 2.7GHz and the memory at 2.62GHz (21Gbps effective). The L2 cache has been consolidated to 36MB, while the TBP has been reduced to 200W. We’re looking at the same shader density as the RTX 3070 with higher clocks, more cache, and faster memory.
Further down the stack, the RTX 4060 Ti will be derived from the AD106 core. We’re looking at a shader count of 4,352 cores paired with 8GB GDDR6 (18Gbps) memory and a 192-bit bus. The L2 cache will be cut down to 32MB, while the TBP is said to be a very impressive 160W. Unfortunately, the performance numbers aren’t looking too good.
Originally tweeted by kopite7kimi (@kopite7kimi) on January 19, 2023.
According to veteran NVIDIA tipster @kopite7kimi, the GeForce RTX 4060 Ti will be roughly as fast as the RTX 3070, which translates to a generational uplift of only 10-20%. This isn’t hard to digest because the 4060 Ti features fewer cores than the 3060 Ti. I’m curious how NVIDIA will pull a $500 SKU as fast as the previous-gen offering at that price point.