NVIDIA has released the official specifications for the GeForce GTX 1660 SUPER and GeForce GTX 1650 SUPER. We already know quite a bit about these cards, but this report clarifies a few things, such as the exact CUDA core count of the GTX 1650 SUPER: 1280, the same as the GTX 1060 series. Also, the GTX 1660 SUPER will only be available through board partners (i.e., it isn’t available as a reference design).
The GTX 1660 SUPER will feature 12nm Turing TU116 graphics processor with 1408 CUDA cores, 88 TMUs, and 48 ROPs. This model will sport 6GB GDDR6 memory across a 192-bit memory bus. This corresponds to 336 GB/s of maximum theoretical bandwidth. That’s even more than GTX 1660 Ti (thanks to 14 Gbps memory chips).
NVIDIA’s [GeForce GTX 1650 SUPER] will have 4GB GDDR6 memory onboard. The card will keep 1650’s 128-bit memory bus and 32 ROP count. The higher specs will come at a cost of 25W higher TDP, as the GTX1650S will now require 100W of power. This model will also be clocked 50-60 MHz higher than the non-Super variant.