The NVIDIA RTX 40, with twice the performance of the RTX 3090?

According to the official roadmap of NVIDIA, They do not have any novelty in terms of new architecture prepared for the remainder of the year, and although they are expected to launch more models of their GPU Ampere, the RTX 40 Series They are still quite a long way off and we will have to wait until at least the end of next year to start hearing from them. However, that does not mean that they are already beginning to talk about the next architecture of those in green, and it seems that the objective will be for the top-of-the-range graphics to have twice the performance of today’s RTX 3090, and it is not to say little.

NVIDIA has provided official data regarding the launch of its new architectures, and in principle after Ampere which is already on the market we will have “Ampere Next” by 2022, followed by “Ampere Next Next” in 2024. Obviously they will not have these names And we all know that, but what is interesting is that they have clearly set a guideline that they will launch a new architecture every two years.

NVIDIA RTX 40 Series, it will be late but it will be worth it

According to the latest rumors (and we have to emphasize that they are mere rumors so they could end up being nothing), the NVIDIA RTX 40 series would not be launched until fourth quarter of 2022 or until the first of 2023, which supposes a certain delay in relation to what we could expect seeing the official roadmap of the company.

Now it is said in the network tricks that the new generation will be manufactured using the TSMC node N5 at 5nm, as the Samsung process is not enough to support such a large core and such high power consumption, and in In terms of performance, it would far exceed everything we have seen so far because, as they say, it would double the performance of the current RTX 3090, the most powerful graphics on the market.

Previous rumors claimed that NVIDIA’s next-gen GPU architecture code would be called Ada Lovelace. The large number of AD102 cores in the new architecture would have up to 12 GPCs (graphics processing group), 72 TPCs (texture processing group), 144 SMs (streaming multiprocessors) and, if each SM continues to have 128 CUDA cores, that means we will be able to see a number of 18,432 Stream Processors on a single chip.

Compared to the 10,752 cores on the Ampere architecture GA102 chip, this is an increase of more than 71%, so the figure of “twice the performance of an RTX 3090” would not really be so far-fetched, but we would already be talking of a graphics card with a power scale that could get out of hand, both in terms of consumption and heat generated as well as in terms of price, as we remember that the price The official RTX 3090 is $ 1,499, already out of reach for most users, so the label that NVIDIA might put on the next-gen top-of-the-line GPU might be far-fetched.

Related Articles