Something to look forward to: Unlike Nvidia’s DLSS and Intel’s XeSS, AMD’s FSR upscaling technology does not rely on AI or machine learning. Instead, it uses a combination of spatial and temporal upscaling, anti-aliasing, and other techniques to enhance image quality from lower-resolution inputs. While AMD’s solution supports a wider range of hardware than its competitors, it typically delivers lower image quality. However, AMD has now confirmed that it is moving forward and has been working toward incorporating AI for some time.
Jack Huynh, senior vice president of AMD’s Computing and Graphics Business Group, recently announced that FidelityFX Super Resolution 4.0 (FSR 4) will use AI. This marks a significant departure from previous versions but potentially opens more questions than the confirmation answers.
Competing upscaling solutions like Nvidia DLSS and Intel XeSS use AI and machine learning to upscale images while minimizing performance impact, though they require specialized hardware not found in game consoles or older GPUs.
In contrast, the latest version of FSR, FSR 3.1, uses spatial and temporal techniques for upscaling and frame generation. It works on nearly any device, but benchmarks show it produces less stable images compared to AI-driven solutions.
Sony’s introduction of AI-based PlayStation Spectral Super Resolution as a key feature of the PlayStation 5 Pro signals a broader industry shift toward AI upscaling – not to mention, the console is based on AMD’s graphics tech. Additionally, the upcoming successor to the Nintendo Switch is expected to support DLSS.
At IFA 2024, Huynh explained that AMD had focused on analytical and filter-based techniques with FSR 2 and 3 to accelerate development, but an AI-based solution has been in the works for around a year.
Interestingly, Huynh’s comments on FSR 4 centered around maximizing battery life on handheld gaming PCs through frame generation. The timing suggests that FSR 4 might be optimized for the AMD Z2 Extreme mobile processor, which the company plans to release early next year. The SoC’s next-generation NPU could be the cornerstone of AMD’s AI upscaling technology.
Huynh did not mention whether FSR 4 would be available on more traditional products like RDNA 4 graphics cards, which are expected to debut in 2025 (though it would only make sense for that to be the case). It also remains unclear whether older AMD GPUs will support the new upscaling technology.
While AMD typically allows its software to work across a wider range of hardware, including competitors’ devices, all AI-based upscaling solutions so far have required some form of dedicated hardware.
Conclusion
AMD’s decision to incorporate AI into its FidelityFX Super Resolution technology marks an important shift in the industry. While AMD’s previous iterations focused on spatial and temporal techniques for upscaling, the move towards AI-driven solutions highlights the company’s commitment to enhancing image quality and performance.
With competitors like Nvidia and Intel already utilizing AI-based upscaling technologies, AMD’s adoption of AI with FSR 4 showcases its willingness to evolve and adapt to meet the demands of modern gaming and graphics processing. The potential benefits of AI-driven upscaling, such as improved image stability and performance optimization, could position AMD as a strong contender in the upscaling technology market.
As AMD continues to innovate and explore the capabilities of AI in its products, consumers can anticipate exciting advancements in image quality and gaming experiences. The integration of AI into FSR 4 opens up new possibilities for AMD’s graphics solutions and sets the stage for future developments in the field of upscaling technology.
