Let’s go back to the topic desktop graphics cards to talk about GeForce RTX4000the next generation of consumer cards NVIDIA based on the new Ada Lovelace GPU architecture, alongside the TSMC a 5 nanometers. At present we know that these products will probably debut after the summer – it is thought in September – promising a marked improvement over the current solutions based on Ampere architecture, all apparently without upsetting the technical approach seen on the GeForce RTX 3000 models.
In our previous article we had discussed the possible features and the GeForce RTX 40, a series that could lead to an important generational leap, not very far from that seen with the transition from Turing to Ampere: we talk about performance more than doubled on the top of the range, reaching up to 2.5 times for the flagship model, the hypothetical GeForce RTX 4090. This clear leap forward apparently “will have a cost” in terms of TDP as well and consumption; he had already talked about it (WHO), but according to the latest rumors of the leaker Greymon55 the scenario that lies ahead is not the best, on the contrary it is worrying. According to what was reported in one of his latest tweets, in fact, the most powerful chip of the lot, Lovelace AD102, will have three TGP configurations (Total Graphic Power), in detail: 450W, 650W and 850WYes, you read that correctly.
Reiterating that this is anything but official informationhowever also confirmed by other sources in the sector, the hypothesis is that these TGP values belong to the most demanding models of the RTX 40 series, in order: GeForce RTX 4080, GeForce RTX 4080 Ti and GeForce RTX 4090. Considering that NVIDIA AD102 is credited with 18,432 Cuda Cores, it is not surprising that the consumption of the top of the range is growing, especially if you think that the much chatted (and postponed) GeForce RTX 3090 Ti it should already sit at 450W with a 10,752 Cuda Core GPU.
Considering that the GeForce RTX 4000 will support the new 16pin PCI-E 5.0 connector (12 + 4), it was expected that the RTX 4090 could reach the peak of 600 watts, but from here to reach 850 watts the jump would be too exaggerated. even for a card capable of doubling the performance of a GeForce RTX 3090. The problems, if we want to call them that, could be different, starting from the power supply required for such a GPU (at least 1,200 / 1,400 watts), arriving at the behavior below the thermal profile. September is still a long way off, NVIDIA remains buttoned up on the subject at the moment, but things should change soon, basically when the first real samples for testers begin to circulate.