Processor designer Arm has developed a new shader-based upscaler for mobile devices with small GPUs but high-resolution screens. What makes this upscaler different from Qualcomm's Game Super Resolution and Nvidia's DLSS Super Resolution is that it is not only open source, but also developed from the results of another open source project: AMD's FSR 2.2 The upscaler.

For better or worse, upscaling has become an expected feature in today's graphics-intensive games. Reducing the rendering resolution can significantly improve performance because fewer pixels are processed, but at the cost of reduced visual fidelity. Upscalers are designed to mitigate this by doing most of the frame rendering at a lower resolution, scaling it back up to the monitor's resolution, and then displaying the frame with the final touches. [Game developers can choose from those that rely on specific hardware in the gaming PC (for example, Nvidia's DLSS Super Resolution requires a GeForce RTX graphics card) to those that are entirely shader-based and require only the general capabilities of the GPU, There are a variety of upscaling methods to choose from.

One of the most widely adopted upscalers is AMD's FidelityFX Super Resolution (FSR), which over the years has gone from a basic spatial upscaler to a complex multi-stage temporal system. While most discrete graphics cards can easily handle shader routines, smaller GPUs, especially those integrated into mobile device processors, are not always capable enough.

To make matters worse, in the case of smartphones, these GPUs have to cope with displays that have a huge number of pixels. Upscaling can really help here, but without enough shader units, complex temporal upscalers are out of the question and simple spatial upscalers must be used.

That is why Arm developed Arm Accuracy Super Resolution (Arm-ASR)-a less complex but still temporal upscaler used in gaming PCs. Basically, the software engineers used AMD's FSR 2.2 as a base, simplified some of the routines, and optimized it for Arm-based low-power processors.

Not only does such a device offer better performance at relatively lower graphics quality, but it also consumes less power. This means less heat generation and longer battery life. [For me, the most important aspect of all this is that the release of the Arm-ASR under the MIT open source license opens the way for game developers to implement it in portable gaming PCs. The APUs in these machines have more powerful GPUs than the typical smartphone, but they must take advantage of upscaling to enjoy the right balance of graphics and performance in big games.

This research may also get Qualcomm's attention for the Snapdragon X chip currently found in all Copilot+ AI PCs. According to Qualcomm, the chip has a GPU that is said to rival any offered by Intel, but word on the street is that the reality does not quite match the marketing claims.

Imagine: using a Qualcomm upscaler developed from Arm-ASR.

Of course, this would not have been possible if the FSR were not open source in the first place, and the fact that anyone can create their own system using it, even if it is not flawless, is a big plus. using AMD's development kit, the DLSS I have seen coders creating mods to replace the FSR, proving once again that open source is a great path to creativity and innovation.

AMD has just released an updated version of the FidelityFX SDK with an improved FSR 3. 1As the frame generation algorithm is completely decoupled from the upscaler, it is only a matter of time before someone makes a mod that implements these for your favorite game. There may also be an Arm-ASR mod for the Copilot+ PC, which would be a great way to get the best out of the game.