NVIDIA’s GeForce RTX 50 are here. Surprise: the RTX 5070 promises to perform as much as the RTX 4090

On this occasion NVIDIA has not caught us off guard. We knew that Jensen Huang, the company’s co-founder and CEO, would introduce the new GeForce RTX 50 during his CES 2025 conference. And here we have them. In addition, this executive has made official the upcoming launch of the four graphics cards that have been leaked in recent weeks: the GeForce RTX 5090, the RTX 5080, the RTX 5070 Ti and the RTX 5070. No surprises so far.

The flagship model, the RTX 5090, is a 92 billion transistor, 318 TFLOPS and 3,352 TOPS beast in the field of artificial intelligence (AI) that in all likelihood will not have competition in the consumer market. Presumably AMD is not going to launch a revision of its Radeon RX 9000 designed to fight with the RTX 5090. In any case, the main contributions of the GeForce RTX 50 are, on the one hand, the Blackwell microarchitecture, and, on the other, the technology DLSS 4 image reconstruction.

NVIDIA GeForce RTX 50: technical specifications

The new NVIDIA GPUs are based on Blackwell and DLSS 4

According to Jensen Huang, Blackwell microarchitecture is the most important innovation in computing graphics since NVIDIA introduced programmable shaders 25 years ago. It is evident that it sweeps home, and soon, when we have the opportunity to analyze the new graphics cards in depth, we will check if this is really the case. Be that as it may, there is no doubt that the GeForce RTX 50 is, on paper, a significant step forward compared to the GeForce RTX 40. As a button shows: according to NVIDIA the RTX 5090 is twice as fast as the RTX 4090 when using image reconstruction using DLSS 4.

In the table that we publish above these lines we have collected the main specifications of the new NVIDIA GPUs, and it is important that we do not overlook that, beyond the numbers, Blackwell arrives with a new generation of Tensor and mapping cores. good heavens. The first are functional hardware units specialized in solving matrix operations that support high parallelization. They have been expressly designed to efficiently execute the operations required by deep learning algorithms and high-performance computing. Tensor cores play an essential role in DLSS technology (Deep Learning Super Sampling).

Another important fact worth looking at is that DLSS 4 will only work on the GeForce RTX 50, and not on previous graphics cards with NVIDIA GPUs, because it requires some of the functional units introduced in the new hardware. graphic. Be that as it may, according to NVIDIA it multiplies performance up to 8 times taking the rendering at the native resolution as a reference thanks to the introduction of Multi Frame Generation technology. This innovation uses AI to generate up to three frames for each rendered frame. Sounds good. We’ll see how it performs when we try it.

In the photograph that we publish just above these lines we can see one of the most interesting data that NVIDIA has made official today: the performance of the GeForce RTX 5070 is comparable to that of the RTX 4090, presumably by resorting to DLSS 4. This feature is important because the RTX 5070 is, at the moment, the cheapest graphics card in the GeForce RTX 50 family. It is no bargain (it will cost 659 euros), but it is evident that it is much more affordable than the RTX 5070 Ti (899 euros), and, above all, than the RTX 5080 (1,190 euros). What remains to be seen is whether it’s worth the financial effort to jump from the RTX 5070 to the RTX 5070 Ti to access the 16GB GDDR7 of VRAM and the additional CUDA, Tensor, and ray tracing cores. We will check it when we analyze them in depth.

NVIDIA GeForce RTX 50: price and availability

These are the prices and release dates of the new NVIDIA graphics cards:

• GeForce RTX 5090: starting at 2,369 euros. January 30 arrives
• GeForce RTX 5080: starting at 1,190 euros. January 30 arrives
• GeForce RTX 5070 Ti: starting at 899 euros. Arrives in February
• GeForce RTX 5070: starting at 659 euros. Arrives in February

rmation | NVIDIA

In Xataka | The next chip revolution is coming. Intel, Samsung, TSMC and AMD are already working on glass substrates

In Xataka | This is the strategy that has kept NVIDIA on top for more than two decades: the innovator’s dilemma