Nvidia will limit factory overclocking on some geforce rtx 20

While working on GPU-Z support for Nvidia's GeForce RTX 20 series graphics cards, Techpowerup has discovered that each GPU model does not have one, but two assigned device IDs, something very unusual and that would be related to a limitation on overclocking.

GeForce RTX 20 has two GPU variants with differences in overclocking

A device ID tells Windows which specific device is installed, so you can select and load the corresponding driver software. Last but not least, Device ID can be used to enable or block certain functions.

We recommend reading our post on How to disable Intel integrated graphics and use the dedicated one from Nvidia

Industry sources have confirmed that Nvidia is creating two device IDs per GPU for Turing, to correspond to two different ASIC codes per GPU model. The Turing -400 variant is designed for use on cards that target the MSRP price, while the 400-A variant is for use on custom designed overclocked cards. Both are the same physical chip, only separated by binning and pricing, which means Nvidia tests all GPUs and classifies them by properties like overclocking potential, energy efficiency, etc.

When a partner uses a Turing -400 GPU variant, factory overclocking is prohibited. Only the most expensive -400-A variants are intended for this scenario. However, both models can be manually overclocked by the user, but the potential for overclocking in the lower model is likely not as high as in the higher rated chips.

All Founders Edition and custom designs seen so far use the same variant of Turing -400-A, which means that device ID is not used to separate the Founders Edition from custom design cards.

Techpowerup font