NVIDIA GeForce RTX 2070 Mobile vs NVIDIA GeForce GTX 1660 Ti (Desktop) vs NVIDIA GeForce GTX 1650 Super (Desktop)

NVIDIA GeForce RTX 2070 Mobile

NVIDIA GeForce GTX 1660 Ti (Desktop)

The Nvidia GeForce GTX 1660 Ti is a powerful, mid-tier, Turing-based, desktop GPU. The GTX 1660 Ti relies on the TU116 chip with 1536 shader units, 48 ROPs, 96 TMUs and a 192-bit-wide memory bus (clocked at 6 GHz, effectively 12 GHz, bandwidth: up to 288 GB/s). The notebook version of the GTX 1660 Ti  is 10 to 20 % slower due to a lower TDP.

The TU116 chip and other Turing-based GPUs are produced using the 12-nm manufacturing process. However, the GeForce GTX 1660 Ti does not offer the latest features such as DLSS (Deep Learning Super Sampling) and Real Time Ray Tracing. Nevertheless, there are some architectural improvements when it comes to floating point and integer performance. These operations can now be carried out simultaneously. Other notable features include Adaptive Shading, unified cache, DisplayPort 1.4, HDMI 2.0b, HDR, Simultaneous Multi-Projection (SMP) as well as H.265 encoding and decoding (PlayReady 3.0).

At Full HD resolution, the GeForce GTX 1660 Ti performs somewhat worse than the GeForce GTX 1070, but significantly better than the GeForce GTX 1060. Therefore, it is best-suited for Full HD gaming. However, it can run less demanding titles in 1440p.

The energy consumption of the GeForce GTX 1660 Ti amounts to 120 watts. The older GeForce GTX 1060 also has a TDP of 120 watts.

NVIDIA GeForce GTX 1650 Super (Desktop)

The Nvidia GeForce GTX 1650 Super for desktops is an entry-level gaming graphics card based on the Turing architecture. The chip, which has been named TU116, was introduced in November 2019. The GTX 1650 series does not offer ray tracing and tensor cores, unlike the more powerful RTX 2000 cards (e.g. RTX 2060). Performance is slightly better than with the GeForce GTX 1650 and comparable with the AMD Radeon RX 5500 XT.

There have also been significant improvements in the Turing architecture in terms of the CUDA cores. Float and INT operations can now be executed simultaneously, the cache has been unified and improved, and Adaptive Shading has been introduced. As a result, Nvidia speaks of an up to 50% higher performance per core compared to Pascal.

The new entry-level card benefits from the advantages of the TU116 chip, since the newer NVENC video encoding delivers more performance compared to Pascal.

Nvidia specifies a power consumption of 100 watts TGP (Total Graphics Power) for the graphics card. The TU11 chip is manufactured by TSMC in a 12 nm FFN process.