.jpg){kind=link}
A few days ago, the alleged specifications of Nvidia"s mid-range 4070 series duo, the GeForce RTX 4070 and the RTX 4070 Ti leaked. Following those, the same leaker and Twitter user kopite7kimi has now revealed the purported specifications of the RTX 4060 Ti as well.
Compared to the higher-end SKUs, the 4060 Ti will reportedly come with GDDDR6 memory instead of GDDR6X. Here are the full purported specs of the RTX 4060 Ti as per the leakster:
- RTX 4060 Ti has a very short reference board.
- The PG190 still uses CEM5 connector.
- AD106-350-A1
- 4352FP32
- 8G 18Gbps GDDR6
- 32M L2
- 220W
Here is a table comparing the specs of the entire RTX 40-series lineup including the alleged specifications of the RTX 4070 Ti, 4070, and 4060 Ti:
| CUDA cores | GPU Die ID | VRAM bus | VRAM capacity & type | VRAM speed & bandwidth | L2 cache | TGP | |
|---|---|---|---|---|---|---|---|
| RTX 4090 | 16,384 | AD102-300-A1 | 384-bit | 24GB GDDR6X | 21Gbps, 1008 GB/s | 72MB | 450W |
| RTX 4080 | 9,728 | AD103-300-A1 | 256-bit | 16GB GDDR6X | 21Gbps, 716.8 GB/s | 64MB | 320W |
| RTX 4070 Ti | 7,680 | AD104-400-A1 | 192-bit | 12GB GDDR6X | 21Gbps, 504.2 GB/s | 48MB | 285W |
| RTX 4070 | 5,888 | AD104-250-A1 | 192-bit | 12GB GDDR6X | 21Gbps, 504.2 GB/s | 36MB | 250W |
| RTX 4060 Ti | 4,352 | AD106-350-A1 | 128-bit | 8GB GDDR6X | 18 Gbps, 288.1 GB/s | 32MB | 220W |
The alleged specifications of these cards, especially the RTX 4070 and RTX 4060 Ti are really interesting as they appear to be cut down in specific areas in accordance with the performance targets set by Nvidia.
The RTX 4070 has a much lower L2 cache compared to the 4070 Ti but it has the same bandwidth. Meanwhile, the 4060 Ti has almost the same L2 cache amount as the 4070, but its memory bandwidth is far less. However, this spec layout makes a lot of sense since the RTX 4070, with its higher 12GB VRAM, will be more geared towards 1440p gaming where higher memory bandwidth is more important. Meanwhile, the 8GB RTX 4060 will be geared towards 1080p gaming and hence, in this case, the relatively higher L2 cache will come in handier than higher overall bandwidth.
Source: kopite7kimi (Twitter)