The GTX 780 was one of the most powerful models in the NVIDIA graphics solutions catalog in 2013 and together with the GTX 780 TI and the GTX TITAN used the true top-of-the-line graphics core of the Kepler architecture, the GK110.
This graphics core had 2,880 shaders, 240 texture units, and 48 raster units, but in the GTX 780, it was reduced to 2,304 shaders, 192 texture units and 48 raster units.
The Kepler architecture has been surpassed by two generations, Maxwell and Pascal, and in general, it has not aged as well as expected, but the most powerful models still have enough “muscle” to move current games in 1080p resolutions and maximum qualities without the problem.
Benchmark has published a video, and we can see how that GTX 780 of 3 GB compares with a GTX 1060 of 3 GB. The first had a launch price of 649 dollars while the second has hovered around 199 dollars, a very large difference that also extends to the performance of both graphics cards.
The GTX 1060 adds 1,152 shaders, 72 texture units and 48 units of the raster, specifications much lower than those of the GTX 780. However, the improvements introduced by the Pascal architecture and its higher working frequencies (GPU at 1.506 MHz1.708 MHz versus the 863 MHz-900 MHz model based on Kepler) allow you to offer a much better performance than that.
By using advanced APIs such as Vulkan we can see that the difference between the GTX 780 and the GTX 1060 of 3 GB becomes enormous (almost double), a detail that we have highlighted because it confirms how bad Kepler takes with advanced APIs like that ( and also with DirectX 12).