It is tradition at NVIDIA to release a new generation of graphics cards every two years. Since that period will soon have passed since the launch of the RTX 30 on the market, it is normal that the expectation for the next generation and for knowing what the first models of the RTX 40 make us bite our nails. The latest information? Well, it could be that those of Jen Hsun Huang had the first representatives of the range already defined. Specifically, the models RTX 4090, RTX 4080 and RTX 4070. What can we expect from the generation gap?
A range of graphics cards is not only its most powerful model, despite the fact that this is its spearhead, but we can also see configurations based on different energy consumption and memory capacities. Apart from using different chips within the same architecture aimed at different screen resolutions. If the RTX 40 stands out in anything compared to previous generations, it will be in the AD102 chip where the graphics cards will have a higher consumption than the current ones when using the new PCIe Gen 5 connector. Which is already currently used by the top-of-the-range model of the current generation, the RTX 4090.
These would be the first models of the NVIDIA RTX 40
Thanks to one of the most reliable information leakers regarding NVIDIA we have been able to know what it will be the initial range of products to be launched on the market and that it will consist of the RTX 4090, RTX 4080 and RTX 4070. Repeating the same nomenclature as in previous generations for the initial launch.
Apparently NVIDIA will deploy three different models and with three different GPUs as the main chip, although they will not use the complete configurations of each of these chips, which we will expect to see in the Ti versions, which will come out at some specific time in the next two years.
- So the RTX 4090 will use a variant of the AD102 with 128 active SMs out of a total of 144. What added to a 450W consumption. At the moment, the initial top of the range will have its VRAM under a 21 Gbps 384-bit bus. And therefore, with a configuration of 24GB.
- Refering to RTX4080your configuration will be 80 SM4 less than the maximum of the chip AD103. Since I didn’t use GDDR6 standard at 18Gbps and its memory bus is smaller as it is 256 bit for a total of 16 GBits performance in theory it would be a little below or on par with the RTX 3080 Ti.
- To finish, the RTX4070 it will be the most modest of all, with the AD104 chip as master of ceremonies. Its configuration will be 56 SM and it will have a bus of only 160 bits, so it will be limited to 10 GB of VRAMalso of the GDDR6 type at 18 Gbps.
What could be its final price?
The fact that none of the models use the full configurations of each of the chips opens up the possibility of Ti models that could appear sometime in the next two years to complement the saga. It should also be noted that there are two chips left under the Ada Lovelace architecture that are the ones with the lowest specifications. We are talking about the AD106 for the potential RTX 4060 and the AD105 for the RTX 4050. We expect the first one in early 2023. The initial range apparently it will have the same starting price that the RTX 3090, RTX 3080 and RTX 3070 had in their day. Whose prices in our country were €1,549, €719 and €519. Although the final price will depend on the economic situation and the conversion from euros to dollars.
As for the consumption, apparently and given the controversy that we mentioned yesterday between NVIDIA and the source manufacturers, it seems that they are going to maintain the consumption of the RTX 30 for the first models of the RTX 40 for the time being.