TechPowerUp has just revealed some very interesting information on NVIDIA’s TU102 GPU, the chip that powers NVIDIA’s upcoming RTX 2080 and RTX 2080 Ti. They noticed that each GPU models has not one, but two device IDs assigned to it. A device ID is the unique identification that tells Windows what device is installed, so it can load the correct driver and tells the driver which commands to send to the chip. The device ID can also be used to enable of lock certain features. Two device IDs per GPU is very unusual, for example the GTX 1080 Ti (custom or reference design) has the device ID of 1B06. NVIDIA has always used just one ID per SKU, no matter if it is custom-design or a Founders Edition card.
It turns out that NVIDIA will be creating two device IDs per GPU for Turing, which corresponds to two different ASIC codes per GPU model. So for example you would have TU102-300 and TU102-300-A for the RTX 2080 Ti. The -300 variant would be designated for cards targeting the MSRP, while the -300-A variant would be for use on custom-designed, overclocked cards. Both are the same physical chip, but separated by binning as well as pricing. So this means that NVIDIA pretests all GPUs and sorts them by overclocking potential, power efficiency, etc.
When a board partner uses a -300 Turing GPU factory overclocking is forbidden. The more expensive -300-A variants are meant for overclocking. Both can be overclocked manually by the user, but it is more than likely that the overclocking potential on the lower binned chips won’t be as high.
Separate device IDs could also prevent someone from buying the cheapest card with reference clocks and then just flashing it with the BIOS from a factory-overclocked card. Think of buying a reference ASUS card and then flashing it to the factory overclocked ROG card.