GeForce's new generation and lots of improvements

At CES in Las Vegas, Nvidia unveiled, along with other products, the new generation of graphics cards, the GeForce RTX 50 Series. In a nutshell, it can be described as pushing the performance and technology capabilities in all areas that Nvidia has been addressing with recent GeForce generations. It offers higher performance, generally better pricing than the last generation, and a host of new and improved technologies like DLSS 4 and Reflex 2.

The highest performance model is the GeForce RTX 5090 with 92 billion transistors, and it can handle 3,352 trillion AI operations per second (TOPS). What’s very interesting about the RTX 5090 itself is the redesign. It has a PCB designed so that there is a free-flowing fin arrangement around the middle of the card. Heated air thus doesn’t pass freely upwards only at the back of the card, but this time also at the flap. But that’s more of a design tweak related to the Founders Edition, with cards from other manufacturers we’ll be more interested in what the hardware underneath the cooler has to offer.

   

Key in achieving higher frame rates in rendering is the new DLSS 4, which adds the ability to generate a higher number of intermediate frames compared to the previous generation DLSS 3. With these new features, Nvidia says this card should outperform the previous RTX 4090 by as much as 100%.

Thanks to AI and more powerful AI computing units, DLSS 4 itself will enable the next generation GeForce to generate up to three intermediate frames between rendered frames. Compared to traditional native resolution rendering, this can offer up to an eightfold performance increase (in terms of frame rates), according to Nvidia.

DLSS 4 is the first deployment in the graphics card space to use a transformer AI model architecture in real-time, enabling better parameter settings and higher image quality.
The Ray Reconstruction and Super Resolution models based on this architecture use twice the number of parameters and four times the computing performance to improve image stability, reduce ghosting, and achieve higher level of detail and better anti-aliasing in games. When the GeForce RTX 50 launches, DLSS 4 should work in 75 games and applications. The following video also mentions that you’ll be able to activate the MFG mode in older games with frame generation using the Nvidia App.

Some of the DLSS 4 improvements achieved with transformer-based models will also be available on older generations of cards. The new version of DLSS is expected to offer higher performance and lower memory requirements. The DigitalFoundry channel got the opportunity to record a preview comparison of both technologies in advance.

However, there is a disadvantage associated with frame generation in the form of having to delay the rendering of the last frame. Similar to DLSS 3, where only the rendering of the last rendered frame was also delayed, the increase in delay should not be significant and should be close to where it is with DLSS 3. In other words, games should behave similarly to how they do with DLSS 3 frame gen in terms of responsiveness.

With frame generation in DLSS 3, Nvidia Reflex technology helped compensate for the delay. With the new DLSS generation, Reflex will also see improvements to help further reduce latency in games. Helping to do this is Frame Warp technology, which is designed to adjust a frame based on mouse movement data just before the frame is rendered to the display.

Reflex 2: a frame more recent than the game render

This probably means that the generated or rendered image will be able to react to mouse movement and move in the appropriate direction just before rendering to the monitor. Thus, perceptually, the frame will react to mouse movement earlier and the position will better match the mouse movement. Surprisingly, this is not a simple camera shift, it can also move objects in the scene in the right direction in addition to the camera movement and fill in dead spots with inpainting, similar to what DLSS does for moving objects. Thus, the scene can be updated to some extent depending on object movement, depth map and other data without having to render the whole scene again. It should even be possible to get a more up-to-date frame even when rendering is limited by CPU performance. This will probably have its pitfalls, as some changes in the scene itself won’t be reflected (for example, it won’t make up a flash of a gunshot when the Warp button is pressed, but will have to get to it from the rendered frame).

The downside is that Frame Warp support will need to be implemented directly into games. First to come with support will be The Finals and Valorant on the GeForce RTX 50, support for older cards should be added later.

First models available at the end of January

So far, Nvidia has unveiled the four highest performance models: the RTX 5090, RTX 5080, RTX 5070 Ti and RTX 5070. Before the unveiling, there was a lot of speculation about their prices. The dark scenarios were probably caused by leaks of the prices of overclocked non-reference models, which will logically be higher.

In the end, except for the fastest model, prices are not significantly higher than the previous generation, quite the opposite. The suggested prices for the new generation of GeForce start at the following amounts:

• GeForce RTX 5090 – 2389 EUR (1999 USD)
• GeForce RTX 5080 – 1199 EUR (999 USD)
• GeForce RTX 5070 Ti – 899 EUR (749 USD)
• GeForce RTX 5070 – 659 EUR (549 USD)

As a reminder, the starting prices for the RTX 40 were as follows:

• GeForce RTX 4090 – 1949 EUR (1599 USD)
• GeForce RTX 4080 – 1469 EUR (1199 USD)
• Geforce RTX 4070 Ti – 899 EUR (799 USD)
• GeForce RTX 4070 – 659 EUR (599 USD)

Please note that dollar prices are excluding VAT, prices in euros include tax.

There is a lot of talk that the higher price of the RTX 5090 is a result of the current lack of competition. However, there’s not much to talk about regarding any price competition between AMD and Nvidia in the high end in recent years. Rather, I would state that the RTX 5090 price is not so much a result of the fact that there is no competition for Nvidia’s high-end models, but rather a result of the current state of supply and demand.

The RTX 4090 sold well above expectations even at 1599 USD, and there’s no reason why Nvidia couldn’t raise the price. The new models have better features and undoubtedly higher production costs, they offer significant added value compared to the RTX 4090, and from that perspective there’s nothing stopping them from slapping on a higher price tag. I’m sure that even at 2000 USD there are enough people for whom such an offer is attractive. It’s hard to expect Nvidia to stick with the same price for a high-end model under these circumstances, let alone reduce it. For others, it still has cheaper models on offer.

Preview of performance compared to last generation

One can assume that the relative comparison of frame rates in the graphs corresponds to reality, but take them with a grain of salt, they are visualisations for marketing comparison, not rigorous work. I’ve added a grid to the images for better performance comparison (the lines are in ten percent increments).

Note that this is not a difference in the raw performance of the graphics cards, but a comparison of frame rates in different rendering modes. The only direct performance comparison is in Far Cry 6 with active hybrid ray tracing, where the performance increase is around 30%, and in A Plague Tale: Requiem, which was only supposed to run on both models in DLSS 3 mode. In other games, the differences are between RTX 40 with DLSS in performance mode and frame generation (i.e. single frame generation) and RTX 50 in 4× multi frame generation mode (multiple frames generation).

The biggest differences are then seen in games with full ray tracing of scenes (path tracing), where multi frame generation has the biggest impact on the differences, and the increase in performance of the compute cores for ray tracing also seems to have a significant impact. It can be assumed that without ray tracing, the differences will be more likely to be somewhere on the level of the results from the first two games.

At CES in Las Vegas, Nvidia unveiled, along with other products, the new generation of graphics cards, the GeForce RTX 50 Series. In a nutshell, it can be described as pushing the performance and technology capabilities in all areas that Nvidia has been addressing with recent GeForce generations. It offers higher performance, generally better pricing than the last generation, and a host of new and improved technologies like DLSS 4 and Reflex 2.

In the case of the GeForce RTX 5090, RTX 5080 and RTX 5070, Nvidia is also offering its own Founders Edition models, while the RTX 5070 Ti will be purely in the realm of non-reference designs from Nvidia’s partners.

GeForce RTX 5090

    

GeForce RTX 5080

    

Nvidia didn’t say much about specific performance or specs at the presentation, but fairly detailed specs are already posted on the GeForce RTX 50 product pages at Nvidia.com. They’re scattered across several tables in various places, so I’ve tried to summarize them:

At CES in Las Vegas, Nvidia unveiled, along with other products, the new generation of graphics cards, the GeForce RTX 50 Series. In a nutshell, it can be described as pushing the performance and technology capabilities in all areas that Nvidia has been addressing with recent GeForce generations. It offers higher performance, generally better pricing than the last generation, and a host of new and improved technologies like DLSS 4 and Reflex 2.

GeForce RTX 5070

    

 

The RTX 5070 looks the same size as the more powerful models due to the proportions of the cooler, but when you compare them by PCIe connector size, it’s similar to the current generation. On top is the RTX 5090, below is the RTX 5070.

The following table summarises how the individual cards of the recent generations differ in terms of technology support.

I guess that’s the most important stuff. Nvidia didn’t say much at the presentation itself, but a lot of material appeared on the web and youtube. I’ve gone through a number of them and tried to summarize and filter the information into this article and not speculate on something Nvidia has already made public. That’s why I’m releasing it a bit later than the gunslingers who only summarized the basic info from the presentation and press conference. I’m sure I haven’t covered everything though, so more info will be added on an ongoing basis. Neural network materials, improved ray reconstruction and other technologies will be worth mentioning, I expect we’ll learn more as the cards go on sale, which should be on January 30 for the RTX 5090 and RTX 5080 and during February 2025 for the RTX 5070 Ti and RTX 5070.

English translation and edit by Jozef Dudáš

• Contents
• GeForce's new generation and lots of improvements
• RTX 5090 and RTX 5080: photos and specs of cards, comparison with RTX 4090 and RTX 4080
• RTX 5070 Ti and RTX 5070: specs, comparison with RTX 4070 Ti Super, RTX 4070 Ti and RTX 4070