Rumors are starting to reach concerning the subsequent RTX 40 fashions that NVIDIA would be engaged on, and they don’t seem to be encouraging. A number of days in the past, the potential traits of the RTX 4070 had been mentioned, with an quantity of CUDA equal to that of the RTX 3070, and now there are rumors that the RTX 4060 Ti would have 4,352 CUDA in comparison with the 4,864 of the RTX 3060 Ti. The turbo frequency would be round 2.6 GHz.
That’s not too promising as a result of in sure conditions the decrease quantity of CUDA could make the RTX 4060 Ti much less performant than the RTX 3060 Ti. Especially in these video games that don’t rely a lot on the frequencies of the CUDA however on their amount. That shouldn’t be misplaced sight of in these rumors.
An enchancment that there would be and that might present additional efficiency in comparison with the 3060 Ti would be the 32 MB stage 2 cache and the 8 GB of GDDR6 at 18 Gb/s. I’m going to be very considerate and I wish to imagine that the AD106 chip on which it would be based mostly would have eight reminiscence controllers for a 256-bit bus and 576 GB / s of reminiscence bandwidth in comparison with 448 GB / s of the 3060 Ti . If there have been 4 controllers, it would be a 128-bit bus and 288 GB/s of bandwidth, which may be very little.
Consumption would be round 220 W, which is greater than the 200 W of the RTX 3060 Ti going from an 8 nm lithograph from Samsung to a 4 nm lithograph from TSMC. Unless, of course, you employ a completely different lithography for the AD106 chip, something is feasible. Even so, these rumors come to say that NVIDIA has gone loopy. It could also be so trying on the value of the RTX 4080.