We've all been there: when your hardware just can't perform adequately on that new shiny game you wish to play. It's frustrating, but with the magic of AI, we might be able to prolong the service life of your gaming rig. I'm talking about image upscaling and frame generation.
What is image scaling and how does it work?
Image scaling is the process of increasing the resolution of an image by adding new pixels. Essentially, your GPU uses machine learning, AI, and algorithms to "guess" what a higher-resolution image would look like and outputs it onto your monitor. This allows you to run a game at a lower resolution while still enjoying a high-resolution gaming experience.
What methods are currently available to the masses?
There are currently a few upscaling technologies on the market. We have DLSS (nVidia), FSR (AMD), XeSS (Intel), and lower-scale implementations like LS1 (LosslessScaling on Steam)
Out of the big three offered by teams red, green, and blue, DLSS (Deep Learning Super Sampling) is widely considered to be the most advanced as nVidia has been heavily developing its AI capabilities. AMD's FSR (FidelityFX Super Resolution) is still a big contender, it's also open source allowing anyone to implement it into their games for free and it's compatible with any GPU but it suffers from a bit more artifacting if you want to pixel peep. XeSS (Xe Super Sampling) is the guy on the block and in many cases it outperforms FSR and it's more affordable as nVidia's pricing has been horrible this past generation, to say the least.
We also have PSSR (PlayStation Spectral Super Resolution) as our fighter on the console side which we will not get into today.
Increasing performance in games.
Upscaling allows your hardware to achieve higher performance in scenarios where it may fall short unassisted. Other than increasing the fidelity or visual quality of your frames by artificially increasing the output resolution. AI upscaling also allows for frame generation. Frame gen is the process of creating an intermediate frame based on data from existing frames. We have seen AI double, triple even quadruple framerate compared to native framerates. Let's say your old computer can run Cyberpunk at 720p low with 30 FPS. That's where these technologies come into play. AI upscaling and frame gen can boost your experience by increasing the gameplay experience to 1080p 60FPS at mid settings.
Limitations.
Though AI upscaling might sound like the saviour of gaming there's a catch and there always is. Upscaling can never compare with native resolution and native frames will always look better than AI-generated ones. It also adds input delay making your gameplay feel not as responsive. Some of this tech is only available on specific hardware like Tensor cores on nVidia. It's also making game developers less inclined to optimize their games as they use AI as a stop-gap solution. AI is the future and sooner than later it's going to be in all areas of computing.