The Nvidia RTX 4060 Ti takes advantage of the fourth version of the Ada Lovelace architecture with an AD106 chip with 22.9 billion transistors. However, the graphics card does not use it to its full capacity since a cluster of “texture processors” is disabled. This brings the number of Cuda units to 4352, when the AD106 has 4608 in its maximum configuration.
For comparison, the RTX 3060 Ti has more Cuda units (4864) and 152 units dedicated to textures, while the RTX 4060 Ti only claims 136. IA increases from 152 to 136, while that of RT Core evolves from 38 to 34 for the raytracing.
The amount of video memory is still fixed at 8 GB. On the RTX 4060 Ti, on the other hand, it is interfaced in 128 bits, against 256 bits for the RTX 3060 Ti. This architectural choice is compensated by a large amount of L2 cache memory on the GPU die (32 MB) and therefore requiring less frequent use of video memory.
The latter is of the GDDR6 type clocked at 2250 MHz. The memory bandwidth is 288 GB/s, equivalent to 554 GB/s according to the manufacturer’s calculations thanks to the L2 cache of the GPU.
Like the other RTX 40s, the RTX 4060 Ti’s GPU is engraved at 5 nm via TSMC’s N4 process, which allows the graphics card to reach 2535 MHz in boost.
On the power side, the GeForce RTX 4060 Ti is announced with an electrical envelope of 160 W, which is 40 W less than the RTX 3060 Ti. According to Nvidia’s measurements, the average power consumption in game would be around 140 W. Again, these are measurements valid for the Founders Edition model. The founder’s partners will offer cards with different frequencies and higher consumption.
