A few hours ago, redditor cavityserach123 received his new card, an RTX 4090, but it turns out that this is one of the revised models with a new GPU. Indeed, the card has an AD102-301-A1, a model normally less expensive to produce.
An RTX 4090 with a revised GPU = less volts?
In itself, this new revision of the chip does not lead to a change in the characteristics of the card. The latter contains 16384 cuda core, 512 TMU and 176 ROP. No, here the change concerns the circuit of the new voltage comparator that can be integrated on the card. This facilitates the production of this model, thus reducing production costs. Well, this cost reduction is felt by the manufacturers, not on the final bill.
Anyway, the redditor realized that his card could take less volts. The maximum voltage went from 1.1V to 1.07V. Now, it remains to be seen if this is generalized to all RTX 4090 or only to Founders Edition models or to his card in particular. Tests will have to be conducted on a larger number of cards to confirm or not the trend.
Finally, concerning the flashing of the bios, it seems that it is not possible to flash this new model with a bios of a 4090 in AD102-300. The fault will come from a change of the sub-ID of the card reference.
In short, this little change is a gift that NVIDIA gives to itself by allowing a small cost reduction. On the user’s side, there is nothing to complain about either in terms of performance or cost.
