Following the GeForce RTX 4070, we’re left with only two more Ada GPU launches. The GeForce RTX 4060 Ti is said to land next month, replacing the RTX 3070 and, to an extent, the 3070 Ti. The RTX 3060 and 3060 Ti should continue to exist alongside, as shown by the recent Steam surveys. Leaked data from the rumor mill indicates that the RTX 4060 Ti may end up being faster than the RTX 3070 Ti while drawing nearly half as much power.
Based on the AD106 graphics core, the GeForce RTX 4060 Ti will feature 4,352 FP32 cores, down from 4,864 on the RTX 3060 Ti and 6,144 on the 3070 Ti. We’re probably looking at a 192-bit bus with GDDR6 (or 6X) memory for an external bandwidth roughly equalling 500GB/s. The RTX 3060 Ti has a wider 256-bit bus but slower GDDR6 memory modules.
Since many are still curious, the RTX 4060 Ti is 8GB, this can change in the future, but as of right now it remains 8GB
According to @T4CFantasy, the editor of the TechPowerUp database, the RTX 4060 Ti will feature an 8GB memory buffer and a core boost of 2.5GHz on FE and up to 2.7GHz on AIB cards. The L2 cache should be cut down to 32MB with a TDP of 160W. This gives us a single-precision (FP32) compute throughput of 22 TFLOPs, a smidge higher than the 21.75 TFLOPs achieved by the 290W RTX 3070 Ti
This doesn’t sound too far-fetched because the 200W RTX 4070 is comparable to the RTX 3080, a 320W card. At the very least, we should see similar numbers (if not greater) in ray-traced workloads with DLSS 3 enabled. Like its 70-class sibling, this would make the RTX 4060 Ti up to 40% faster than its predecessor with a price tag of $499.