(Image credit- MMORPG.com)
Starting at $399 for the Founders Edition and reference-clocked variants, the Nvidia GeForce RTX 4060 Ti introduces genuine mainstream pricing to the Ada Lovelace architecture and RTX 40-series GPUs.
The 128-bit memory interface and the 8GB of VRAM are the two biggest potential drawbacks, which is unfortunate. The 16GB variant from Nvidia, which is scheduled for availability in July, may offer a solution to the capacity issue, but it won’t fix any issues with the memory interface.
Are the RTX 4060 Ti graphics cards among the best? This mostly relies on your gaming habits, the number of games that support DLSS 3, and if you’re ready to sacrifice latency for AI-interpolated FPS.
At 1080p, the RTX 4060 Ti edges out the RTX 3070 in terms of native performance in our GPU benchmarks hierarchy (which will be updated later today), but at 1440p and 4K, it lags the RTX 3060 Ti.
The good news is that the RTX 4060 Ti is generally quicker than the RTX 3060 Ti from the previous generation while consuming less power. Additionally, it supports recent additions to Ada, such as DLSS 3 Frame Generation, SER, DMM, and OMM. The bad news is that, generally, it hardly outperforms its predecessor, and design choices from years ago are undoubtedly at work.
Although the table is packed, the first column is the most important. The RTX 4060 Ti utilizes Nvidia’s brand-new AD106 GPU, which is, coincidentally, the same chip used in the RTX 4070 Laptop GPU. The block diagram for the 4060 Ti and the AD106 chip, which we’ll discuss shortly, is also shown below.
As a result, there are fewer GPU cores than in the RTX 3060 Ti, but Nvidia makes up for this with core clocks that are much higher — 2535 MHz boost against 1665 MHz.
Real-world clocks will, as usual, be faster than those numbers, but in terms of theoretical computing from the shaders and tensor cores, the 4060 Ti achieves 22.1 teraflops compared to 16.2 teraflops for FP32 and 177 teraflops compared to 130 teraflops for FP16 (with sparsity).
The 4060 Ti has tensor cores that support FP8 mode as well, thus it has the potential to deliver 353 teraflops if AI applications decide to support it.
We have AMD’s RX 6750 XT with 12GB and Intel’s Arc A770 with 16GB when comparing the competition based on somewhat similar costs. In terms of ray tracing performance and AI workloads, Nvidia can match or even outperform such cards—winning the latter by default as it’s frequently the only GPU option supported. Rasterization performance, on the other hand, will be a lot closer and a more interesting comparison point.
The performance will be greatly influenced by memory size and bandwidth. Most games running at 1080p shouldn’t have any trouble with 8GB of VRAM, however, 1440p and especially 4K could cause issues.
Also read: GTX 1660 Ti vs. GeForce RTX 3050: Two GPUs for two distinct uses
Recent discussions concerning the increased VRAM needs of 4K gaming are relevant here. Because of its limited VRAM capacity and bandwidth, Nvidia isn’t advertising the RTX 4060 Ti as a 1440p or 4K gaming solution. That’s odd because the 3060 Ti and 3070 both aimed for 1440p two years ago.