Twitter leaker Kopite7kimi claims the RTX 4070 Ti will have an AD104 GPU and be so powerful that it can trade blows with the current Ampere flagship, the RTX 3090 Ti.
The AD104 GPU, built on the TSMC N4 process node, is said to feature 7,680 CUDA cores or 60 SM units, 30 billion transistors, and up to 160 ROPs. The RTX 4070 Ti is believed to have 48 MB of L2 cache and 12GB of GDDR6X memory at 21 Gbps on a 192-bit bus, giving it a maximum 504 GB/s bandwidth.
The downside is that the RTX 4070 Ti will suck up 400W TDP which is not that far from the RTX 3090 Ti's 450W.
The RTX 3090 Ti is built on Samsung's 8nm process, while the RTX 4070 Ti has the advantage of 1.7 billion extra transistors and a more modern process node. The Ampere card is ahead in terms of CUDA cores (10,752), GDDR6X memory (24GB, also at 21Gbps), and memory bus (384-bit).
Not only could the RTX 4070 Ti also be a lot cheaper than the current-gen card, which Nvidia recently reduced from its $1,999 MSRP to $1,599.
In addition to the latest claims of a mid-range product being able to match a current-gen flagship, the RTX 4090 could be twice as fast as the RTX 3090 and 66 per cent faster than the RTX 3090 Ti.