Nvidia has received a fair amount of criticism for its power hungry high-end GeForce RTX 40-series graphics cards based on the Ada Lovelace architecture. According to renowned hardware leaker @kopite7kimi, Nvidia is now looking to make its mainstream GeForce RTX 4060 Ti part a lot more power efficient.
As ever with leaks, take the news with a pinch until real world benchmarks and tests can be undertaken.
After correction, 160W. https://t.co/thAbqh0NdqJanuary 18, 2023
According to kopite7kimi, the GeForce RTX 4060 Ti will have a thermal graphics power of up to 160W instead of 220W reported initially. Meanwhile, general specifications of the product are set to remain generally the same: the board will be powered by the AD106 graphics processing unit with 4352 CUDA cores and accompanied by 8GB of 18 Gbps GDDR6 memory connected using a 128-bit interface.
Right now, it is unclear how the new TGP affects GPU frequencies and performance of the GeForce RTX 4060 Ti and AD106 processor. With 4352 CUDA cores, the GeForce RTX 4060 Ti is already significantly behind rumored specifications of the GeForce RTX 4070, which allegedly has 5888 CUDA cores. If AD106's GPU clocks are dropped significantly from those planned initially, performance gap between GeForce RTX 4070 and RTX 4060 Ti will be significant.
| Row 0 - Cell 0 | GPU | FP32 CUDA Cores | Memory Configuration | TBP |
| GeForce RTX 4090 | AD102 | 16384 | 24GB 384-bit 21 GT/s GDDR6X | 450W |
| GeForce RTX 4080 | AD103 | 9728 | 16GB 256-bit 22.4 GT/s GDDR6X | 320W |
| GeForce RTX 4070 Ti | AD104 | 7680 | 12GB 192-bit 21 GT/s GDDR6X | 285W |
| GeForce RTX 4070 | AD104 | ?5888? | 12GB 192-bit 21 GT/s GDDR6X | ?250W? |
| GeForce RTX 4060 Ti | AD106 | ?4352? | 8GB 128-bit 18 GT/s GDDR6 | ?160W? |
With 8GB of 128-bit GDDR6 memory, Nvidia's GeForce RTX 4060 Ti could offer a memory bandwidth that is 43% lower compared to that of the GeForce RTX 4070, so the gap between two products in memory bandwidth-dependent workloads (e.g., in high resolutions and/or with antialiasing enabled) may be huge. Meanwhile, the new product will also feature a 36% lower memory bandwidth than its direct predecessor, the GeForce RTX 3060 Ti with GDDR6 memory. Of course, we would expect 32MB of L2 cache to somewhat reduce bandwidth limitations, but only tests will reveal whether this is the case.
Meanwhile, carrying only 8GB of memory is another peculiarity that can also be seen as a limitation. We already felt the 8GB of memory on the GeForce RTX 3060 Ti, 3070, and 3070 Ti was becoming a limiting factor and those are outgoing products launched in 2020 – 2021. Despite all these factors, the GeForce RTX 4060 Ti is still likely to join the ranks of the best graphics cards as it will easily outperform its predecessor.
With a 160W TGP, the GeForce RTX 4060 Ti could position itself as a viable graphics solution for small form factor gaming PCs. The board is rumored to feature a very short reference design, so expect Nvidia's partners to offer dual-slot boards for Mini-ITX machines and perhaps even single-slot variants for thin PCs.
Keep in mind that we are talking about specifications leaked by an unofficial source, so take them with a pinch of salt, though it is plausible that Nvidia decided to build an energy-efficient yet powerful GPU to better compete for compact gaming PCs.
