Cheaper RTX 20-series cards will not come factory overclocked?
btarunr of TechPowerUp found something very curious about the Turing GPUs. Each Turing GPU has two device IDs. He noticed this issue when he was working on GPU-Z support for the latest NVIDIA GeForce RTX 20-series cards.
Apparently, the two device IDs correspond to two different ASIC codes per GPU. For example, the GeForce RTX 2080 Ti will have TU102-300-A and TU102-300 GPUs, with the former used in the factory overclocked cards from AIBs and also the Founders Edition cards from NVIDIA themselves.
The latter GPUs without the -A suffix are forbidden from being sold overclocked from the factory. It seems that they do support manual overclocking, but considering that NVIDIA goes through the trouble of making two device IDs per GPU, it wouldn’t be too outlandish to assume limited headroom on the binned cards.
What do you think? Will this make you reconsider getting the latest GeForce RTX cards?
Pokdepinion: These GPUs would actually be great for the compact cards intended for ITX PCs. Cheaper, and since the cooling would limit overclocking anyway, the limited headroom is much less of an issue.