How exactly does Nvidia technology work?

In addition to ray tracing, Nvidia primarily advertises its existing graphics cards with a new functionality: DLSS. But what exactly is all this? What does the new graphics functionality bring? And when should I enable DLSS? GIGA enlighten you.

DLSS: What is it – and what does graphics technology bring?

The DLSS acronym stands for “Deep Learning Super Sampling” and is one of the nvidia Upgraded graphics technology to ensure higher gaming performance. More precisely, it is a type UpgradeThe technology by which the image is calculated at a lower resolution and then upgraded by artificial intelligence algorithms. Nvidia uses a neural network created specifically for this purpose.

However, for the AI ​​to know how to scale up the corresponding game scenes to make them look as good as possible, the algorithm must first be “trained”. Therefore, the network is fed a large amount of high-resolution images. Calculations in real time on your computer can only be done using this data. Ago The computation of individual images by a neural network is less computationally intensive From the original view, eventually more images can be generated. So your game runs more smoothly without sacrificing graphic quality.

Still not convinced? The following video impressively demonstrates why DLSS can be worthwhile:

DLSS 3: The most important differences with DLSS 2.0

Well, it’s getting a little confusing now. Strictly speaking, DLSS is 3 meaning Not the successor to DLSS 2.0, although the name might suggest so. Instead, DLSS 3 is a mixture of several technologies: Super Resolution (DLSS 2.0), Nvidia Reflex, and Frame Generation. The latter is part of the Ada Lovelace architecture and currently only works with corresponding RTX 4000 series graphics cards. This is perhaps the biggest catch.

So what exactly are all these (new) technologies? The brief explanation is as follows:

• Ultra HD: This upgrade technology is also known as DLSS 2.0 and therefore is not really new. In combination with frame generation, the Theoretically quadrupled FPS in games.
• Nvidia Reflex: All CPU and GPU related calculations are synced with System access time To reduce. Since frame creation brings with it a slight increase in latency, Nvidia Reflex provides adequate compensation.
• Frame Generation: This technology is a completely new feature that uses the algorithms and “optical flow accelerators” found in the GPU. Additional frames created. This effectively doubles the FPS.

As good as this technology sounds, the new RTX cards didn’t really convince me:

What graphics cards support DLSS?

So far, DLSS is only available to owners of an RTX graphics card – and for good reason. Nvidia has been using only the necessary tensor cores in its cards since this generation. The more powerful the graphics card, the more computing cores with AI capabilities on the board. The following Nvidia GeForce GPUs currently support DLSS:

If you want to take advantage of all the functions of the new DLSS-3 standard, you need a current-generation RTX card – eg RTX 4080 or 4090.

By the way, you do not need an internet connection to use DLSS. Computed game profiles for the graphics card supercomputer are provided via a framework, that is, a type of programming framework, called NGX. This will be in Install the graphics card driver installed at the same time.

What games support DLSS?

RTX cards have now been around for several generations, thus, DLSS is also supported by more and more games. Below you will find a small selection of the current addresses where DLSS operates:

One A complete list of all compatible games and appsYou can find it directly on Nvidia.

Besides DLSS, ray tracing is another technology that will make your games look better:

While DLSS 2.0 is now relatively widespread, DLSS 3 is currently only supported by a select few games. The Complete list of all DLLS-3 compatible games It can be found as usual at Nvidia. But here’s a little more to expect:

When is DLSS worth using – and when not?

DLSS is most effective when you need the highest resolution possible and the original frame rate is less than 60 frames per second (frames per second). In other words: If the graphics card has become a real bottleneck, then using DLSS is worth it. If you also want to take advantage of ray tracing, you can’t avoid DLSS. According to Nvidia, the performance of some games can be doubled or even tripled by enabling DLSS 2.0. Thanks to the frame generation found in DLSS 3, performance can even be quadrupled in some cases. However, the actual increase in performance also depends on many other factors, which is why you should always treat this information with a healthy degree of skepticism.

However, there are also scenarios where using DLSS might not be helpful at all. DLSS needs a fixed amount of time for each image to be computed to access the neural network data. Sometimes this can take longer than the original calculation of the image – provided the graphic requirements (settings, resolution or quality) are not too high. So if the obstacle isn’t the graphics card, but the processor or some other component, DLSS usually doesn’t offer any real added value.

DLSS without an RTX graphics card: AMD and Intel make it possible

DLSS may be the most widely used technology, but the competition never sleeps – and in the end, players benefit the most from it. AMD’s answer is called “FidelityFX Super Resolution” and basically works in a very similar way. However, the main difference is that the technology is not limited to internal Radeon graphics cards, but also works with GPUs from other manufacturers. With the release of FSR 2.1, the technology has also been improved in a number of points and will be introduced in more games in the future.

Intel is now also in the graphics card business and also has an upgrade technology in its portfolio with “Xe Super Sampling” (XeSS for short). Similar to AMD’s FSR, the approach here is that the software can be used independently of the respective graphics card manufacturer. However, only if the technology is supported by the game in question. Intel is currently in last place here, since the standard only works with relatively few addresses so far.