By pure logic, the next graphics card to be launched by NVIDIA should be the RTX 4070, which would be based on a version with fewer active cores of the AD104 chip used in the RTX 4070 Ti and would even use the same base cards. However, the latest information tells us about three variants of the RTX 4070. What would they be and how would each of them be different?
Little by little the mid-range of the RTX 40 is preparing its market launch and it is that although the new generation has been able to show its strength, it has done so at a price much higher than most of between us mortals can afford economically. And what do we mean today by this point? Well, a price between 500 and 600 dollars. Perhaps that’s why NVIDIA considered launching not one, but up to three variants of the RTX 4070, so we’d be faced with an exercise from the folks at Jen Hsen Huang like the one they tried with the 12GB RTX 4080.
What would be the three variants of the RTX 4070?
The information comes to us from GIGABYTE, more precisely from the certification process of its graphics cards, where three variants of the RTX 4070 have been seen, the differences between them? Well, we would be talking about a 10 GB model, another 12 GB and finally a 16 GB. At the moment the information was that it would use the same chip as the RTX 4070 Ti, but with 5888 of the so-called active CUDA cores, for a total of 46 real or active SM cores. Unlike the 60 cores of the AD104 chip in its full configuration.
The interesting thing is that such a chip has a 192-bit bus and no more. Hence the possibility of placing 6 GDDR6X chips of 2 GB each to reach 12 GB, so the version with this capacity corresponded to what we expected. As for the 10GB fan, all it takes is removing a memory chip and cutting out a few more cores. Even if the great unknown is in the 16 GB model, which is only possible with a 128-bit bus in clamshell mode, two chips per interface instead of one, or under a 256-bit. Although in the latter case we would really be talking about the AD103 chip, that of the RTX 4080, which makes all the information less strange.
Why is the 16GB model an anomaly?
Normally we tend to associate that the more memory a graphics card has, the bigger its bus, however, it is possible to place two chips to expand capacity without doubling the bus. The result? In this case, a 16GB memory card, but less capable than 10GB and 12GB cards. The good thing is that it will not be limited by lack of memory, but it will be limited in terms of processing. At this time, this has not been confirmed, but we find such a policy confusing for the end user.
Anyway, if we start from the fact that the AD107 chip of the RTX 4060 has 36 SM cores in total and the 12 GB 4070 has 46, then the other two versions, the 10 GB version with a 160-bit bus and the 16GB with a 128-bit bus should fall somewhere in between the two configurations, but we don’t know the full specs at this time.