Nvidia unveils new GeForce RTX 50 graphics cards, which double the FPS performance of the previous generation

• GeForce RTX 5090: 2,329 euros
• GeForce RTX 5080: 1,169 euros
• GeForce RTX 5070 Ti: 879 euros
• GeForce RTX 5070: 649 euros

Original news: The CES is fertile ground for hyperbole. Only the fact that the fair is held in Las Vegas seems to favor the most exaggerated proclamations, but what Nvidia has done this morning has few precedents. And it is that, with its new RTX 50 graphics cardspromises to double the performance of the previous generation in FPS.

Based on the Blackwell architecture and equipped with up to 32 GB of GDDR7 VRAM, the company’s new consumer GPUs follow to the letter the leaks that have been precipitating in recent weeks, with a total of 21,760 CUDA units in the case of the GeForce RTX 5090 and a range that arrives well articulated from day one, starting from a GeForce RTX 5070 that acts as a stepping stone. access with 6,144 CUDA units and 12 GB of GDDR7.

The keys to understanding this performance increase are three: first, the Blackwell architecture is by far the most powerful ever developed by Nvidia, incorporating improvements to the Tensor and RT cores, not to mention GDDR7 memory; second, the TDP has taken a good jump, reaching 575 TDP in the case of the RTX 5090; and thirdly, DLSS 4 technology allows you to generate three frames “synthetics” for each traditionally rendered one.

The performance data provided by Nvidia is truly impressive, at least on paper. If we are to trust the official information, not only does the RTX 5090 double the power of the outgoing RTX 4090, but exactly the same happens with the RTX 5080, 5070 Ti and 5070 when compared to their direct predecessors. It will be necessary to see to what extent they are sustained in practice and how much improvement remains without DLSS.

Despite this, and curiously, they are relatively compact cards. Although the increase in TDP (which was already known) portends the worst, the RTX 5090 is a conventional dual-slot card, so it should have no problems finding a place in the vast majority of towers… as long as they are equipped with a 1,000 watt power supply, which is what Nvidia recommends for this card.

Other notable improvements that come thanks to Blackwell are compatibility with PCIe Gen5 and DisplayPort 2.1b UHBR20, which in theory enables these cards to display images at 8K and 165 Hz.

Prices for Spain have not yet been disclosed, but in the United States they will be as follows:

• GeForce RTX 5090: $1,999
• GeForce RTX 5080: $999
• GeForce RTX 5070 Ti: $749
• GeForce RTX 5070: $549

GeForce RTX 50 also for laptops

On the other hand, the RTX 50 will also have its own interpretation for laptops, both in Max-Q equipment, characterized by being relatively thin and light, and in larger machines.

Nvidia has not provided specific performance figures or benchmarks for these graphics cards, but notes that Max-Q laptops, with a minimum thickness of 14.9 mm, will also double the FPS figures compared to the previous generation, all while while providing an improvement in consumption close to 40%.

As usually happens in this type of GPU, the RTX 50 range for laptops has a series of optimizations and changes in relation to desktop cards, starting with a lower number of CUDA cores and the use of specific memory configurations with smaller modules. consumption. The range is summarized as follows:

• GeForce RTX 5090 Laptop: 10,496 CUDA cores, 1,824 AI TOPS and 24 GB of GDDR7
• GeForce RTX 5080 Laptop: 7,680 CUDA cores, 1,334 AI TOPS and 16 GB of GDDR7
• GeForce RTX 5070 Ti Laptop: 5,888 CUDA cores, 992 AI TOPS and 12 GB of GDDR7
• GeForce RTX 5070 Laptop: 4,608 CUDA cores, 798 AI TOPS and 8 GB of GDDR7

The first laptops equipped with Blackwell graphics will be available starting in March.