Nvidia RTX cards are part of the next generation
Technology is always evolving and graphics hardware is no exception. Each year, cards become more powerful and offer a new range of graphics tricks.
You’ll see a word salad in the visual settings of PC games. It contains MSAA and FXAA, as well as WWJD. Okay, that’s not the last one.
You can now enable DLSS if you’re the lucky owner of a new Nvidia GeForce card. It stands for Deep Learning SUPER Sampling. This feature is a major part of Nvidia RTX cards’ next-generation hardware features.
These cards are the only ones that have the hardware required to run DLSS at the time of writing
- RTX 2060
- RTX 2060 Super
- RTX 2070
- RTX 2070 Super
- RTX 2080
- RTX 2080 SUPER
- RTX 2080 Ti
This hardware is called a ” Tensor” core. Each model has a different number.
Tensor cores can be used to accelerate machine-learning tasks. DLSS is one example. This means that DLSS is not used and the card will remain idle. If DLSS is not available, you’re not using your GPU to its full potential. However, it remains off.
However, there’s more to it than this. We need to briefly explore some related concepts to understand the value of DLSS.
Quick Detour into Internal Resolutions and Upscaling
Modern monitors and TVs use what is called a “native resolution. This means that each screen has a certain number of physical pixels. The screen must be scaled up or down if the image displayed is not the native resolution.
If you export an HD image to a 4K display, it will look very blocky and jagged. It’s like zooming in on a digital photograph too much. Although HD video is sharper than native 4K footage, it looks great on a 4K TV. The TV’s “upscale” is a piece that processes lower resolution images and makes them acceptable.
Problem is, that the quality of upscaling hardware can vary widely between models and brands. This is why GPUs are often equipped with their own scaling technology.
Pro consoles are designed to output to 4K displays. They present the image with native 4K resolution, which prevents any display scaling. The game developers have full control over the final image quality.
Most console games don’t render at native 4K resolution. These games have a lower “internal”, which places less strain on the GPU. The image is then scaled to the highest resolution screen by using the console’s internal scaling technology.
DLSS, in effect, is a complex method that renders a PC gaming program at a lower resolution than its native resolution and then uses DLSS technology for upscaled output on the connected display. This theoretically results in a significant increase in performance.
Although it sounds a lot like 4K consoles, DLSS under the hood is truly something special. Thanks to “deep learning”.
What’s the “Deep Learning” Bit About?
Deep learning uses a virtual neural net to aid in machine learning. This is a digital simulation of the way your brain’s neurons learn and solve complex problems.
It is the technology that allows computers to recognize faces, and robots to understand and navigate the environment around them. It is also responsible for recent deepfakes. This is the secret sauce to DLSS.
Training neural networks are basically showing them examples of how something should look. You can show the net millions of faces to help it recognize a face. If the net learns the lesson correctly, it can be shown any image that has a face in it and it will instantly recognize it.
Nvidia has trained their deep learning software using extremely high-resolution images taken from games that support DLSS. The neural network then learns how the game should look when rendered with supercomputer-level graphics performance.
The computer then uses the lower resolution frame to “imagine” how it would look if your computer had rendered this scene. You’re not the only one who thinks this sounds like black magic.
When to Use DLSS
First, DLSS can only be used in games that support it. This list is growing rapidly, thankfully. Every title has its own requirements regarding DLSS. For example, rendering at a minimum resolution is required because that’s what our neural net was trained on.
Nvidia’s big brain doesn’t stop learning. The DLSS feature on the card will continue to get updates. This will allow for per-title support and better quality.
Eyeballing the results is the best way to determine if DLSS should be used in your games. It is easier to compare it to anti-aliasing or traditional upscaling. Performance is another important factor. DLSS is a great option if you want to achieve 60 frames per second but cannot do so.
DLSS may slow down frame rates if you have high frame rates. Because the tensor cores take a set amount of time to process each frame, this is why DLSS can slow things down. They can’t process high frame rates at the moment because they aren’t fast enough.
DLSS can be most effective when used with a high-resolution display (e.g. You can use DLSS with high-resolution displays (e.g., 4K, ultrawide, or 1440p resolutions) at a target frame rate of around 60 frames per second. It is also beneficial for activating ray tracing, the main party trick of the RTX card. DLSS can be used to offset the performance losses of ray tracing and produce spectacular results.
This is all you need to know before you decide whether to use DLSS. This technology is constantly changing so don’t be surprised if the results don’t impress you.