An early engineering sample of the NVIDIA GeForce GTX 2070 has been revealed, featuring a core count lower than the original model. The GeForce "GTX" branding has been consistent from the GeForce 10 series all the way to the GeForce 16 series GPUs. Following these, all subsequent products have been labeled under the GeForce "RTX" series. However, it seems that internally NVIDIA may have considered labeling the GTX 20 series instead of the RTX 20 series, as evidenced by the appearance of such a model online.
The image shared by @unnatural__log depicts an unusual NVIDIA GeForce GTX 2070 GPU, reportedly an engineering sample, which explains why it lacks the "RTX" label.
In terms of design, the NVIDIA GeForce GTX 2070 engineering sample adopts the same design as the GeForce RTX 2070. It features the same dual-fan heatsink and dual-slot design, but the label on the side reads "GeForce GTX," with no mention of "GTX 2070" or "RTX 2070" on the front. There are a few missing labels on the card, but other than that, there are no major differences. The display ports include the same HDMI, 2x DP, USB Type-C VirtualLink port, and DVI output. The card should use a standard 8-pin interface, as it predates NVIDIA's switch to 12-pin or 16-pin interfaces.
The PCB of the NVIDIA GeForce GTX 2070 engineering sample also matches the reference design, but its specifications are the major change. While the RTX 2070 has 2304 cores, the GTX 2070 ES GPU has 2176 cores, based on the TU106-400A-A1 GPU, with device ID "10DE-1F07 - 10DE-12AD." This matches the core configuration used by the RTX 2060 SUPER, which uses the TU106-410-A1 GPU.
According to tests conducted by users, the NVIDIA GeForce GTX 2070 engineering sample GPU was flashed with the RTX 2070's "400A BIOS." This BIOS doesn't change the device ID or unlock additional cores but provides some extra power to the GPU core for manual tweaking. Through overclocking, the GPU's performance eventually reached 95% of the RTX 2070's, representing a 16% boost over the default configuration. Considering that modern GPU clock speeds often surpass their listed "boost clock" speeds by a significant margin, this is quite a commendable improvement.
