Do you have a GeForce RTX 40 SUPER card and you don't use it? You're missing out on a lot

Graphics cards today have much more to offer than just pure computing power capable of generating graphics in standard rasterization. Many additional techniques significantly affect not only the image quality, but also the smoothness of the game. It's worth using them.

Just a few years ago, when buying a graphics card, only one thing mattered – computing power. We were all looking for models with the best possible price-performance ratio. Of course, we could not forget about energy consumption or work culture, but they were and in many cases still are secondary issues. After all, what gamers care most about is as many frames per second as possible.

Today that has changed a bit. Not only do we look more often at the performance of individual models and energy consumption, but also at other parameters, apart from the efficiency of standard rasterization. The capabilities of graphics cards in solutions such as Ray Tracing or image scaling techniques, such as Deep Learning Super Sampling, have become equally important. NVIDIA is the market leader in this matter. If you don't use these techniques yet, believe me, you are missing out.

GeForce RTX 40 Super graphics cards

In January this year, new GeForce RTX 40 Super series graphics cards debuted. In total, three models appeared on store shelves: GeForce RTX 4070 Super, GeForce RTX 4070 Ti Super and GeForce RTX 4080 Super. These are models that have been significantly improved compared to traditional designs. What exactly has changed about them?

Let's start with the cheapest model, GeForce RTX 4070 Super. In it, NVIDIA primarily increased the number of CUDA units from 5,888 in the regular model to 7,168 in the refreshed version. This also resulted in more ROP, TMU and RT cores. This alone gives a solid boost in terms of performance, but the Greens have additionally increased the amount of second-level cache by as much as 33%. up to 48 MB. Contrary to appearances, this is a very important change that has a real impact on the capabilities of the entire structure.

The last change was to increase the base clock to 1980 MHz, but in Boost mode it remained at 2475 MHz. However, we are talking about reference models, because there are also overclocked designs on sale, e.g. ASUS ROG Strix GeForce RTX 4070 Super OC Edition. The clock increases up to 2640 MHz in default mode and 2670 MHz in OC mode. This translates into even more frames per second and better gameplay smoothness. Reference models already provide an improvement of 15 percent, but in the case of ASUS it will be even more.

Next on the list is the GeForce RTX 4070 Ti Super model. There have been even more changes to it. The increase in CUDA units may not be as impressive as in the lower model, because they increased from 7680 to 8448, but we have more VRAM here (16 GB instead of 12 GB), which additionally use a 256-bit bus, not 192-bit. This resulted in an increase in throughput from 504 to 672.3 GB/s.

The clocks of the GeForce RTX 4070 Ti Super model are 2340 MHz base and 2610 MHz in Boost. However, again it is worth remembering about non-reference models, which can be significantly improved in this respect. An example is the ASUS ROG Strix GeForce RTX 4070 Ti Super OC Edition variant, in which the clock frequency is up to 2700 MHz in Boost mode.

The least has changed when we compare the GeForce RTX 4080 and GeForce RTX 4080 Super models. The number of CUDA units increased from 9,728 to 10,240, and the VRAM speed increased to 23 Gbps. The change was small, but resulted in an increase in bandwidth by 19.5 GB/s to 736.3 GB/s. The clocks have also increased to 2295 MHz base (an increase of 90 MHz) and 2550 MHz in Boost (an increase of 45 MHz). Such clocks can also be found in the ASUS ROG Strix GeForce RTX 4080 Super version, but again we can cite the example of overclocked models, e.g. ASUS ROG Strix GeForce RTX 4080 Super OC Edition, which has a frequency of 2640 MHz (default mode) and 2670 MHz (OC mode ).

But why is it worth choosing GeForce RTX 40 Super models? There are many reasons for this, but the most important is not only very good performance, top work culture and relatively low energy consumption (compared to the competition), but also how perfectly they support solutions that we did not even think about a few years ago. What I mean here is the performance in Ray Tracing, as well as the NVIDIA DLSS technique with all its positives. I'll explain what it's about.

NVIDIA DLSS, or regret not using it

Let's start with NVIDIA DLSS, because in my opinion it is a much more interesting and very future-proof solution. I realize that Deep Learning Super Sampling has as many supporters as opponents. I can't understand the latter. They often argue that it is cheating and generating artificial images and frames.

This justification does not appeal to me. Mainly because game developers have always been cheating us and will continue to cheat us. Many different methods are used to generate graphics in games, including the so-called smear frames, i.e. blurry frames that look bad and unnatural when stopped, but when the animation is in motion, we don't notice any problem.

It's similar with NVIDIA DLSS. Yes, the image is scaled and the frames are generated by artificial intelligence, and if we look at each of them closely, we can probably see minimal differences. However, during normal gameplay it is completely invisible. However, we get a much smoother gameplay that is much more pleasing to the eye. What's more, it happens practically for free. In some cases, you can even talk about improving the image quality, e.g. sharpness.

And how does this happen? Today, NVIDIA DLSS is actually several solutions that work closely together. It started with image scaling techniques. This allows the graphics card to natively generate a lower resolution image and thus produce more frames per second. At the same time, machine learning algorithms improve and scale each individual frame to a higher resolution. In this way, the card is loaded as if we were playing in 720p, and we see a 1080p image on the screen.

Over time, NVIDIA added the already mentioned generated frames, which appeared in DLSS version 3.0. Here again, artificial intelligence is helpful – it analyzes the image and inserts additional images between natively generated images, thus improving the smoothness. The tests here are clear. Take, for example, the GeForce RTX 4080 Super card and the Cyberpunk 2077 game. After activating Ray Tracing and at the highest graphics settings, it generates about 30 frames per second by default. After enabling DLSS with Frame Generation, this value increases to 73 fps. That's more than twice as much! Free, without player involvement and in a way where it is difficult to notice any difference.

You do not believe? If you haven't checked out how NVIDIA DLSS works yet, there's no point in resisting any longer. If this solution is available in a given game, it is worth enabling it, along with frame generation. You will see that there was no point in refusing. You can enable Deep Learning Super Sampling in many games, including Diablo IV, Horizon Forbidden West, Dragon's Dogma 2 and LEGO Fortnite.

Anyway, let's see how much DLSS brings in Horizon Forbidden West on a computer with a GeForce RTX 4080 Super card:

Game settings	Average frame rate
1080p	150.1
1080p + DLSS 2	175.5
1080p + DLSS 3.5 (Frame Generation)	200.8
4K	74.1
4K + DLSS 2	100.8
4K + DLSS 3.5 (Frame Generation)	123.3

Ray Tracing for visual connoisseurs

We also cannot forget about Ray Tracing. Many people may not realize this, but ray tracing is nothing new. This technique has long been used, among others, in cinema. The difference is that when creating a movie, the image does not have to be generated in real time. A short scene can be processed on a computer for several hours. However, in games, the image is constantly processed by graphics cards, and previously they were not efficient enough to cope with such an advanced task.

This changed only with the premiere of GeForce RTX 20 series graphics cards, which NVIDIA equipped with special RT cores. These are responsible for ray tracing calculations, thanks to which we can now observe extremely realistic graphics in which appropriate light plays a huge role. This solution is still used today, although it has been improved with each subsequent generation and today there are no better ray tracing cards than the GeForce RTX 40 series, with particular emphasis on the refreshed models from the Super line.

Thanks to Ray Tracing, graphics cards can track thousands of light rays. Their behavior is calculated in real time, such as passing through a transparent glass, reflection from a mirror or only partial reflection from the surface of a frosted cup. In this way, GeForce RTX 40 Super models can faithfully reproduce the lighting conditions of a given scene, bringing us even closer to photorealistic graphics.

It is worth mentioning DLSS 3.5 here, because in the latest version NVIDIA focused on improving image quality, with ray tracing enabled. A technique called Ray Reconstruction is responsible for this. It is worth knowing that graphics cards are currently capable of measuring several samples for each pixel, and since some pixels have no information about light at all, noise is created as a result.

Ray Reconstruction is designed to remove this noise and further improve image quality. An appropriately trained neural network is used for this purpose. Fed with thousands of images, it is able to remove noise from each frame generated by the GPU in a fraction of a second. However, while DLSS 3 requires GeForce RTX 40 series graphics cards, including the previously mentioned ASUS ROG models, owners of older models can also use ray reconstruction.

A beautiful future

Graphics cards take full advantage of the capabilities of artificial intelligence and machine learning algorithms. There is no point in defending yourself against it. Both NVIDIA DLSS and real-time Ray Tracing support were created to make games even prettier, even more immersive and even more engaging, and at the same time as smooth and pleasing to the eye as possible. Today, it is difficult to really point out any reasonable arguments against using these solutions on GeForce RTX 40 Super cards. It just doesn't make sense. If you're still not doing this, you're really missing out. It's best to see for yourself and enable both techniques in your favorite games now.

We also cannot forget that GeForce RTX graphics cards are equipped with dedicated Tensor cores that can be used not only in games. They are perfect for all tasks related to artificial intelligence. Just install NVIDIA Broadcast to dramatically improve audio and camera image quality, or Chat with RTX to create a personalized, local GPT language model. The GeForce RTX 40 Super card is ready for this too.

Sponsored article commissioned by ASUS.