According to pulsestacks on April 20th, the NVIDIA GeForce RTX 2070 was released on October 17, 2018, with the FE Founders Edition priced at 4999 yuan.
As shown in the image released by @unnatural__log, when NVIDIA initially developed the 20xx series graphics cards, they did not anticipate the later adoption of the RTX naming scheme. Therefore, this "GTX 2070" graphics card still retains the traditional GTX prefix.
The "GTX 2070" graphics card, as shown, features the same design as the reference RTX 2070 card, with dual axial fans for cooling, a dual-slot, dual-fan setup, and a unified metal shroud. It retains the iconic side-lit branding and offers HDMI, dual DP, USB-C VirtualLink, and DVI output ports, utilizing a standard 8-pin connector.
In terms of specifications, the "GTX 2070" is based on the TU106-400A-A1 GPU, with a device ID of "10DE-1F07 - 10DE-12AD," slightly differing from the RTX 2070.
Compared to the RTX 2070 with 2304 CUDA cores, the GTX 2070 has only 2176 cores, resembling more closely the RTX 2060 Super's TU106-410-A1 GPU. This suggests NVIDIA may have initially developed it as an "RTX 2060 Super" solution before opting for a higher-spec GPU for the RTX 2070.
According to tests, this GTX 2070 can be flashed with the RTX 2070 "400A BIOS" firmware. However, even after flashing this BIOS, the device ID won't change, nor will it unlock additional cores. It will, however, unlock some extra feature functionalities.
After manual overclocking, the GPU ultimately achieved about 95% of the performance of an RTX 2070, which is a 16% improvement over the default configuration. At this point, the frequency has far surpassed the "peak speed" officially provided.
As far as pulsestacks knows, the NVIDIA GeForce RTX 20 series adopts the new "Turing" architecture and 12nm process, supporting both "RT ray tracing" and AI technologies. For instance, the RTX 2070 comes with 8GB of 256-bit GDDR6 memory onboard.
Beyond its robust specs, the highlight of the "Turing" architecture is its support for "RT ray tracing technology." This innovation accelerates the processing of light propagation in three-dimensional environments. It seamlessly integrates real-time ray tracing, artificial intelligence, and programmable shading techniques. The speed at which it handles rays is 25 times faster than Pascal architecture products. Moreover, when compared to using the CPU as a node processor for final frame rendering, using the GPU speeds up the process by nearly 30 times, resulting in more realistic and stunning lighting effects.
