Tech InDepth: Understanding Nvidia’s DLSS and what’s new with DLSS 3.0

If you’re into PC gaming and have played any of the recent games, ‘DLSS’ is a setting you’ve probably come across a few times. The graphic setting is quickly shaping up to be one of the most important graphics settings that could potentially change a user’s experience of a game without any new hardware to their system.
But what exactly is Nvidia’s DLSS tech? How does it do what it does to improve performance and what is new with the recently announced DLSS 3.0 that Nvidia showcased alongside its new RTX 40-series cards? Here’s all you need to know in today’s edition of Tech InDepth.
What is DLSS?
Deep Learning Super Sampling, or simply put DLSS is Nvidia’s own tech that powers most of its newer graphics cards. The tech uses machine learning to upscale games in real-time, using AI models to output high-res scans of in-game elements.
What this means is if your graphics card supports DLSS, you can turn on the option and configure it to play games at higher resolutions than what your GPU was natively capable of, or to better the performance of an exing resolution, which is the more common use-case.
For instance, if your system is great for playing a particular title at 1080p, but outputs unsatisfactory frame rates while playing the same title at 1440p (2K resolution), your system’s hardware limits you to playing at 1080p for an enjoyable, smooth experience.
However, if you add DLSS into the equation, the same system would be able to generate 1080p frames and use DLSS to upscale this in real-time to 2K resolutions before sending the same to your display. This would enable you to play at 1440p while your GPU is working to generate 1080p frames.

DLSS-powered 1440p will not likely be as good as true 1440p resolution, but will enable more users to play the title at the higher resolution. This is useful when you have hardware limitations, and don’t want to upgrade multiple components to play at a higher resolution.
DLSS is powered Nvidia’s Tensor cores in its newer GPUs. The Tensor cores are dedicated cores in the Nvidia’s RTX-series GPUs, found in the RTX 20, 30 and the new 40-series graphics cards. While the RTX 20-series cards used first-generation Tensor cores for what was essentially DLSS 1.0, second-gen Tensor cores since the RTX 30-series have largely improved upon the technology.
What about DLSS 3.0?
With Nvidia’s new RTX 40-series cards that were launched earlier this month, the company brings DLSS 3.0 in the picture, the most advanced version of the tech so far. DLSS further improves upon the upscale abilities of the new Tensor cores and provides more frames and better image quality with more details.
Nvidia showed off the new DLSS iteration in a recent video where Microsoft Flight Simulator, a popular, graphic-intensive title can be seen running with DLSS off and DLSS On (with ray-tracing) side- side. Check it out below.

The numbers shown in the video suggest that we can expect almost twice the framerate with DLSS enabled compared to native resolutions, although more realic comparisons should be found on the web in a few months across various games when the RTX 40-series cards become accessible to more testers and general users.

Unfortunately, DLSS 3.0 has been limited to the new RTX 40-series cards, which would come as bad news for RTX 30-series users. Until recently, prices for RTX series GPUs, especially the newer 30-series ones were super high and except for a few lucky users who were able to buy the graphics cards at MRP, most users who bought an RTX 30-series GPU paid a premium for it.
This is why these cards being limited to DLSS 2.0 will not make gamers very happy. However, the good news is that DLSS 2.0 is not going anywhere for now and you can continue to use it to better the output of your games. Moreover, the RTX 40-series cards are also super expensive at their official prices, with the RTX 4090 coming in at Rs 1,55,000, and not everyone would want to spend that much for better resolution upscaling on their games.