“Reviving Pixels: How AI Upscaling Transformed Gaming for the Better”

Even for someone like me, a devoted enthusiast of graphics card hardware, it took a while to get accustomed to the concept of AI resolution upscaling.

However, this innovative technology has been adopted in recent years by nearly all prominent GPU manufacturers, with Nvidia at the forefront with their DLSS technology, while others like AMD with FSR, Intel with XeSS, and even Sony with PSSR on the PlayStation 5 Pro have followed suit.

And I’m now prepared to declare: real-time AI upscaling technology has rendered native resolution graphics irrelevant in gaming. Continue reading to discover what AI upscaling involves, why it is groundbreaking, and what implications it holds for all the upcoming games you will be enjoying in the future.

Related: The finest graphics cards for every financial plan

What is AI resolution upscaling?

AI resolution upscaling (often referred to simply as AI upscaling) occurs when a game renders its frames at a resolution that is below your display’s native resolution, then employs AI image processing methods to enhance that rendered frame back to native resolution. Thanks to contemporary machine learning, these upscaled images appear remarkably close.

Not every AI upscaling feature is created equal, of course. For instance, Nvidia’s DLSS (Deep Learning Super Sampling) performs best with its RTX GPUs, while AMD’s FSR (FidelityFX Super Resolution) and Intel’s XeSS (Xe Super Sampling) focus on broader compatibility at the expense of performance. In certain scenarios, the disparities can be exceptionally significant.

Nvidia’s DLSS is generally recognized as the top performer, especially when combined with its frame generation technology. Running on the tensor cores of an RTX GPU, frame generation examines two frames and utilizes AI to produce an additional frame in between, yielding higher frame rates and smoother graphics in CPU-limited games. Nvidia Reflex subsequently reduces input latency that may be experienced during this kind of frame creation.

For example, imagine you’re playing Cyberpunk 2077 at Ultra settings with 4K resolution and Path Tracing. Even the Nvidia GeForce RTX 4090 would find it challenging to achieve playable frame rates. However, with Nvidia DLSS 3.5 and frame generation, the performance sees a substantial boost without significantly compromising visual quality. It’s the best of all scenarios.

On the other hand, AMD has not fully integrated AI into its FSR 3 technology as of yet (but intends to do so in the near term) and has its own variant of frame generation known as Fluid Motion Frames. Intel’s AI upscaling with XeSS 2, which we recently evaluated in our review of the Intel Arc B580, also incorporates AI. Intel’s technology additionally contains Xe Low Latency (XeLL), similar to Nvidia Reflex.

Why AI upscaling is groundbreaking

No matter the potency of a GPU, it will eventually encounter limitations. It can only perform to a certain extent without significantly increasing its cost and power requirements. This is why AI upscaling is creating such a disruption.

There was a time when Nvidia promoted Scalable Link Interface (SLI) technology, which allowed multiple GPUs to be linked for simultaneous processing to produce higher frame rates. This was, naturally, costly and cumbersome, requiring ample space… hence the discontinuation of SLI.

Conversely, individual high-performance GPUs like the GeForce RTX 4090 can only achieve so much independently. This is where AI upscaling comes into play. If executed effectively, AI upscaling creates the opportunity for enhanced performance — even with limited hardware. For instance, the previously mentioned Intel Arc B580 is a budget-level GPU that can deliver substantial performance owing to AI upscaling features.

Nonetheless, that is not necessarily the groundbreaking aspect. Enhancing the capabilities of a weaker GPU is beneficial, and boosting a powerful GPU’s abilities is exciting, but the true advance comes from manufacturers being able to develop low-power devices that can achieve significantly more.

Think about a portable gaming device like Valve’s Steam Deck. It’s beautifully designed, yet its compact build restricts the amount of hardware it can incorporate — this entails overcoming numerous challenges such as thermal management, engineering constraints, etc. These kinds of gadgets cannot simply insert a more robust GPU; they require support from alternative sources. (In the case of the Steam Deck, Valve utilizes AMD’s FSR to enhance game visuals.)

This concept can be extended to a variety of similar gadgets, including game consoles. Consoles are frequently designed to particular criteria to manage costs, thermal output, and dimensions. This explains why the PlayStation 5 Pro now features PSSR (PlayStation Spectral Super Resolution), its proprietary AI-driven upscaling technology that enables it to exceed its expected capabilities.

With PSSR, the device can learn and evolve progressively — which is quite groundbreaking. This not only implies that the gadget can perform admirably but can also enhance itself over time. Should Sony collaborate with AMD to devise future advancements in AI upscaling to further boost console efficiency, the future of gaming could indeed become captivating.

Why native graphics are nearly obsolete

One potential drawback of AI upscaling is that some developers have begun to adopt it as an excuse for inadequate optimization. As AI upscaling becomes more commonplace and as additional developers grow accustomed to integrating it into their games, it is likely we will witness a decline in game performance when these features are not present.

Additionally, hardware manufacturers like Nvidia and AMD will persist in enhancing their AI upscaling technologies because they aim for their tools to be preferred by developers. With the numerous advantages derived from AI upscaling, players who rely on native resolution are facing a dual challenge: not only are they experiencing inferior performance, but that disparity in performance is going to continue to expand. Some might argue this is merely the cycle of GPU evolution.

Nonetheless, despite the superiority of Nvidia’s technologies, AMD remains an important contender for one vital reason: its standalone desktop GPUs may hold a smaller market share than those of Nvidia, yet AMD continues to thrive in the console space. AMD continues to be the backbone powering the PlayStation, Xbox, and Steam Deck, and the company’s advancements in AI upscaling will only further enhance these devices as they move away from native graphics gaming.

Of course, AI upscaling is not without its flaws. Residual artifacts and unusual performance problems can still emerge, and the efficacy of AI upscaling can vary significantly from one developer to another. However, as adoption rises and more individuals depend on it, native graphics will likely become a relic of the past.

How to begin utilizing AI upscaling immediately

At the time of writing, the optimal way to experience AI upscaling is through a high-performance GPU, such as those found in Nvidia’s GeForce RTX 40 Series. If your finances are limited, you can still take advantage of AI upscaling affordably with Intel’s newly launched Arc B580 GPU.

Numerous games already support AI upscaling, which you will recognize as Nvidia DLSS, AMD FSR, or Intel XeSS within the display/graphics options of those games. You can select the quality level that suits you, which will influence performance outcomes.

Generally, AI upscaling will enable you to run a game at a higher frame rate than your hardware GPU would normally manage.allow at native resolution. For instance, Microsoft Flight Simulator running at 4K offers approximately 65 FPS natively but achieves 122 FPS with DLSS 3 and frame generation — a notable enhancement with minimal drawbacks that might go unnoticed.

With the advent of next-gen GPUs, AI upscaling is set to improve significantly. The Nvidia GeForce RTX 5090 GPUs are expected to feature more sophisticated AI capabilities, including a refined iteration of DLSS.

Ray tracing and path tracing have emerged prominently in gaming over recent years, with games like Cyberpunk 2077 and Alan Wake 2 famously showcasing these technologies. They present stunning visuals, yet even the most powerful GPUs struggle under their demands without the aid of AI upscaling.

Nonetheless, Nvidia is likely to continue advancing ray tracing in the forthcoming generation, with upgraded RT cores in its GeForce RTX 50 Series GPUs. This will render AI upscaling increasingly attractive to both developers and gamers, enabling graphics that are currently unattainable.

Given all the aforementioned factors, I anticipate that the next wave of GPUs from Nvidia, AMD, and Intel will mark the widespread embrace of AI upscaling, leading to the phasing out of purely native graphics.

Further reading: Don’t pay extra for factory-overclocked GPUs!