With just a few days to go before the presentation of the RTX 40, the next NVIDIA graphics cards, the partial specifications of the RTX 4090, RTX 4080 16 GB and RTX 4080 12 GB have already been leaked. So the only thing that would wait for us is to know the changes in architecture and the novelties in that aspect that the Ada Lovelace architecture brings with respect to the current one. What models have been leaked and how is each of them presented?
For NVIDIA there is a titanic task in the coming months, the launch of its new generation of graphics cards for PC gaming is compounded by the problem of the definitive end of the mining bubble and the flooding in the market of RTX 30 models That is why they have decided to start with the most powerful models, for the moment, in an almost repeated launch with three card models appearing in the high range.
These are the models that debut the new generation of NVIDIA
In total three would be the models that NVIDIA will launch to release its RTX 40 cards, repeating again what we have already seen in the last two generations of graphics cards. As a curiosity, there is a name change at the last minute of one of the models, as happened a couple of years ago where the rumored RTX 3080 Ti evolved into the RTX 3090 that we all know. In this case, it is the RTX 4070 that has made the leap to be called the 12 GB RTX 4080 . Hopefully this change is not also reflected in the final price, which we still do not know, but seeing the picture, we better cross our fingers.
RTX 4090
The most powerful is the RTX 4090, this will have a wide variety of different models and will appear headed by the AD-102-300 chip with 16,384 of the so-called CUDA “cores” , that is, 32-bit floating point units. What 128 SM means if there have been no configuration changes that we don’t know about. By the way, one thing we don’t know yet is its base speed, but we do know the Boost speed, 2520 MHz . Below the 3000 MHz or 3 GHz that was speculated a few weeks ago. Its memory is 21 Gbps GDDR6X for a bandwidth > 1 TB/s. Although what is striking is the margin of its TGP, which apparently will go from 450 W to 660 W.
RTX 4080 16GB
In the middle part of the table will be the RTX 4080 with the AD-103-300 chip and with 9,728 CUDA “cores”, and therefore 76 SM, a Boost speed of 2505 MHz and 16 GB of GDDR6X memory , but at 23 Gbp s. That is, 736 GB/s of bandwidth . Its TGP is 340 W , but in the design of the card it can reach 516 W. The fact that it does not reach the maximum possible configuration with the AD103 chip makes us think that there is a variant called AD103-450 that would be the RTX 4080 Ti and would come out much later.
With said TGP, the initial model would not need the 12 + 4-pin connector to work, since it would not differ in that aspect with respect to current graphics. On the other hand, future models that approach the maximum of the consumption and temperature capabilities of this model could carry it. In any case, everything seems to indicate that it is the limit for the card plate and its components to support the eventual RTX 4080 Ti. Which makes us think that it could be the same case with the model that we have previously mentioned regarding an eventual RTX 4090 Ti. In any case, we would only have to wait a few days.
RTX 4080 12GB
Despite its name, this is not a cut in terms of the amount of memory of the 16 GB RTX 4080, since the 12 GB model uses a different chip, but obviously with the same architecture, the AD-104- 400 with 7680 CUDA “cores”, 60 SM, and a 2610 MHz Boost speed. Its capacity is 12 GB and it uses GDDR6X memory at 12 Gbps , which translates into a bandwidth of 504 GB/s . As for the consumption of the card, it moves between 285 W and 366 W.
