AMD told us its plan for how FSR will eventually kill DLSS

As an Amazon Associate, we earn from qualifying purchases. TweakTown may also earn commissions from other affiliate partners at no extra cost to you.

"In the long run, open source wins," Frank Azor says at the end of a lengthy conversation. At Computex 2025, we got the chance to sit down with AMD's Consumer and Gaming Marketing chief to discuss the evolution of FSR 4 and the latest FSR 'Redstone' announcement.

6

FSR 'Redstone' introduces new AI rendering technologies that will rival DLSS 4's impressive AI-powered Ray Reconstruction, Neural Radiance Cache, and Frame Generation for real-time Path Tracing in games. It's due for release in the second half of 2025.

"Look at G-Sync and FreeSync," Frank Azor concluded. "It took a while, but FreeSync won. And that's par for the course; it just takes some time." This is where our conversation concluded. He used the open-source nature and widespread adoption of FreeSync as a reference point for how FSR would one day kill DLSS as the go-to solution for Super Resolution and Path Tracing technologies. Of course, there's a lot more to it than that.

6

AMD is well aware that it is playing catch-up to NVIDIA, so to speak, regarding these features. This is where my discussion with Frank Azor on FSR 4 began: about getting these technologies and the AI-powered FSR 4 in more games to match DLSS 4's widespread adoption.

FSR killing DLSS Phase One - Open Source and Easy to Implement

"Everyone likes the number, but not all games are created equal," Frank Azor tells me. "Not all people are playing the same games, so the number of games (FSR 4 is available in) is not nearly as important as what games you support. If you want to do a competitive analysis, you have to look further than just the 60 versus 125 games number. We went from 30 to 60 in six weeks. The adoption is rapid, and an incredible amount of groundwork has been laid with FSR 1, 2, 3, and 3.1."

FSR 4 Super Resolution is widely viewed as a new beginning for the technology, based on its dramatic improvements to image fidelity. However, it is exclusive to RDNA 4 GPUs and requires powerful AI hardware, so it's a good thing the Radeon RX 9000 Series launch has been its most successful to date.

6

"As demand for our graphics cards grows, so does the install base," Frank Azor explains. "And the easier we make FSR to be adopted, the more rapid the adoption. We've made FSR 4 very easy to adopt for developers. I would say that it's the easiest machine learning upscaling solution that's out there on the market today."

"We made FSR ubiquitous, open source, easy to adopt, and applicable to not just our graphics cards but also our competitors' graphics cards," Frank Azor continues. "Our competition has almost the exact opposite strategy to that."

It's harder to create an open-source solution than something proprietary. And when you arrive late to the party, the adoption rate accelerates. If something is easier to implement, open source, and ubiquitous, the logic is that a game developer would choose that option over something else. "Unless there's incentives to do otherwise," Frank Azor adds, referring to what we assume are development deals for using a specific technology.

FSR killing DLSS Phase Two - PlayStation and Project Amethyst

Project Amethyst is a collaboration between AMD and Sony to develop machine learning solutions to accelerate performance on PlayStation consoles. PlayStation 5 Pro's AI-based PSSR solution, or PlayStation Spectral Super Resolution solution for upscaling, is the precursor to this Project Amethyst becoming a reality and FSR 4 being integrated into the PlayStation ecosystem.

"If you're developing for PlayStation and other platforms, where are you going to allocate your upscaling resources?" Frank Azor asks rhetorically. "The technology that will benefit you across PC and multiple consoles, or apply your resources to a proprietary technology that's not applicable to the console version, and only the PC version. I think that answer is pretty clear."

6

One key advantage AMD has up its sleeve when it comes to the adoption of FSR 4 is that it has been co-developed with Sony for PlayStation and, by extension, future consoles like the upcoming PlayStation 6. Although Frank Azor did the expected thing by not responding to any mention of the PS6 and shifting the conversation, Project Amethyst changes the landscape for how the company traditionally works with hardware partners in the console space.

"If you look at the relationship with our partners, historically, they've been very hardware-centric," Frank Azor says. "Hey, we're building this chip, and we're going to power this console. The dialogue has evolved in 2025 to be upscaling and software-centric. I think that gives a good indication of how the future is evolving between the companies."

FSR killing DLSS Phase Three - Mission Accomplished?

Of course, a few roadblocks are standing in the way of FSR killing DLSS. FSR 4 is currently exclusive to RDNA 4 gaming GPUs due to its AI hardware requirement, and there's no guarantee it will work in its current state on the current PlayStation 5 Pro hardware.

This means that the Sony drawcard might not come into effect until the PS6 arrives. So, as of now, DLSS 4 is available to run on exponentially more GPU hardware than RDNA 4 because it works on older GeForce RTX GPUs, which cover the majority of PC gamers. A number that we're guessing even eclipses the current PS5 install base of 77 million. On top of that, NVIDIA will continue to develop newer versions of DLSS and introduce more AI-powered rendering technologies like it is already doing with the GeForce RTX 50 series.

6

Also, as of now, FSR 4 is not open-source like previous versions or available to run on multiple generations of GeForce RTX GPUs. This isn't to say that FSR killing DLSS would never happen, but to highlight that the road might be longer than you'd think. The good news is that AMD is working toward making FSR 4 open-source.

"Based on how (FSR 4) is currently programmed, it won't run on anything else as it's honed in on the RDNA 4 architecture," Fran Azor confirms. "But, it's going to be open source. We are going to make it available out there. It's not impossible. I don't think it's impossible. It just requires work, and the work has to get done. We're an open-source company, and that's our strategy."

"In the long run, open source wins."