For a few years we have known that NVIDIA’s plans include offering the most powerful graphics cards with each generation, although this greatly affects the price they may have and other really important factors such as consumption. But it seems that this does not present a problem for the company, since they want to find a way to increase the capabilities of their cards at any cost, and that would also involve developing a model superior to the flagship they have now, the RTX 5090and the brand could be planning to create a more advanced model of this GPU since a prototype of your version You.
One of the main reasons why many people were waiting for both the presentation and the official launch of the new NVIDIA graphics cards is related to the power they are capable of offering. Although they are extremely expensive models, it is impossible to deny that they have hardware that we can truly consider the most advanced in terms of performance. The specifications with which the architecture GPUs arrive Blackwell They far exceed people’s expectations, such as the bandwidth of the RTX 5090, which is a 77% increase compared to the RTX 4090.
NVIDIA could have in mind creating a more powerful version of the RTX 5090
There are some hardware components that manage to stand out for one reason or another, the RTX 5090 has clearly managed to do so because of the specifications it has, but also because of what it costs and even what it can consume. This model considered the most powerful of the current generation has some key aspects such as 32 GB of VRAM, 1792 GB/s of bandwidth and a TGP of 575 W, but all this could be completely overshadowed by a model that is A prototype has been discovered and it could represent the true flagship of this generation.
This has to do with an alleged leak that has been shared through the Chiphell forums, and a user has indicated that NVIDIA I would be preparing a PCB complete, very different from the compact version that uses the graph RTX 5090. In this case, it stands out that it uses a GB202-200-A1 GPU instead of the shortened version that the normal model would incorporate, and this chip would arrive with 13% more CUDA cores, going from offering 21760 to 24576, with a configuration of the clock that would be between 2100/2514 MHz.
It should also be noted that this model also uses the maximum power that GDDR7 is capable of offering, the basic version has 1792 GB/s because the memory is configured at a lower speed than this standard can achieve, so this The supposedly improved 5090 could exceed 2 TB/s by incorporating 32 Gbps memory. But if something should be highlighted above all this, it would be the increase in TGP, since the normal model already consumes too much, 575 W, so it needs a 12V-2×6 connector, but this model could need two of these connectors, since their TGP would be 800 W.
