The GeForce RTX 2070 used to be one of the best graphics cards. Many don’t know that it was internally known as the GeForce GTX 2070 before Nvidia adopted the RTX branding for products with RT cores while keeping the GTX moniker for those lacking them. X user Jiacheng Liu has shared his experience with one of these rare engineering samples.
The GeForce GTX 2070 Founders Edition looks almost identical to the GeForce RTX 2070 Founders Edition. The main difference is that the former carries the GeForce GTX branding on the side of the graphics card, hinting that it’s an early engineering sample before the transition to the RTX branding. The GeForce GTX 2070 uses the same TU106-400A-A1 silicon as the GeForce RTX 2070, but the CUDA core count differs.
The GeForce GTX 2070 sports 2,176 CUDA cores, 128 less than the GeForce RTX 2070. We’re looking at two disabled SMs on the GeForce GTX 2070 or a GeForce RTX 2060 Super. The die is different as the GeForce RTX 2060 Super employs the TU106-410-A1 silicon. Liu successfully flashed the GeForce GTX 2070’s vBIOS with a 400A BIOS from a GeForce RTX 2070. Logically, it didn’t unlock extra cores or anything of that sort. However, it raised the power level, allowing for manual overclocking headroom.
Liu pushed the GeForce GTX 2070 to the edge and achieved a performance of around 95% of that of a GeForce RTX 2070. The overclock represents a 16% uplift over the stock performance. In reality, 128 CUDA cores aren’t a lot, so it was surprising that the enthusiast got the GeForce GTX 2070 within a hairline of the GeForce RTX 2070.
The GeForce GTX 2070 leverages the same Founders Edition cooler as the GeForce RTX 2070. It obviously didn’t carry the RTX 2070 branding yet. However, the single 8-pin PCIe power connector and the mix of a DVI port, one HDMI 2.0 port, two DisplayPort 1.4a outputs, and the USB Type-C port are still present even on this prototype.
The emergence of the GeForce GTX 2070 engineering sample insinuates that Nvidia originally planned to outfit the GeForce RTX 2070 with fewer CUDA cores. Although Nvidia didn’t go through with it, the chipmaker didn’t throw the idea in the trash since the CUDA configuration later ended up in the GeForce RTX 2060 Super. Turing is in the past, but it’s always intriguing to see the early prototypes and how they fare with the retail products.