AMD Expands FSR 4.1 ML-Powered Upscaling to Legacy Graphics Cards

The wait for advanced upscaling on older hardware is finally over. AMD has officially announced that it is bringing FSR 4.1, its latest machine learning-powered upscaling technology, to the RX 7000 and RX 6000 series graphics cards. This move marks a significant shift from the company’s previous strategy, which restricted the newer, smarter upscaling features exclusively to the latest RX 9000 series.

Jack Huynh, AMD’s GM of Computing and Graphics, confirmed the rollout schedule via social media, detailing a phased approach for existing users. The update timeline reflects the architectural differences between AMD’s recent generations:

  • RX 7000 Series (RDNA 3): FSR 4.1 will arrive in July 2026.
  • RX 6000 Series (RDNA 2): Users will need to wait until early 2027.

This announcement resolves a long-standing frustration for the PC gaming community. While AMD’s newer RX 9000 series (RDNA 4) currently hosts the full suite of ML-powered technologies under the "FSR Redstone" umbrella, older cards have been left behind in terms of native support.

Bridging the Hardware Gap

FSR 4.1 harnesses machine learning to deliver image quality comparable to Nvidia’s DLSS. However, the technical requirements for this technology differ significantly between hardware generations. The version currently available on RDNA 4 cards is specifically designed for FP8 (8-bit floating point) support, leveraging the 2nd Generation AI Accelerators found in the RX 9000 series.

In contrast, the 1st Generation AI Accelerators in RDNA 3 support INT8 (8-bit integer) but lack FP8 capabilities. Similarly, the compute units in the RDNA 2 architecture of the RX 6000 series also rely on INT8.

While the absence of FP8 support might seem like a dealbreaker, it does not preclude FSR 4.1 from functioning on these older cards. This potential was previously demonstrated by Sony’s PSSR (PlayStation Spectral Super Resolution) upscaling technology. As confirmed by Mark Cerny, PSSR is based on AMD’s technology but utilizes INT8 instead of FP8. Although PSSR and FSR 4.1 are not identical, this proof of concept showed that ML-powered upscaling could run effectively on older AMD hardware.

Modders Beat the Manufacturer

The delay in official support became a sore spot for gamers, especially after the open-source nature of AMD’s technology allowed the community to step in. AMD published the source code for FSR 4 INT8 on GitHub, which was quickly compiled into a DLL by modders.

These community-driven patches enabled FSR 4 functionality on older graphics cards long before AMD could officially release it. By expanding native support to the RX 7000 and RX 6000 series, AMD is effectively righting this wrong, albeit on a slower timeline.

"I’m grateful to our fans," Huynh stated regarding the expansion. "Your enthusiasm and ideas inspire us to keep pushing gaming forward."

Clarifying the FSR Ecosystem

It is crucial to distinguish between FSR 4.1 Upscaling and the broader FSR Redstone suite. The upcoming updates pertain specifically to the upscaling component, not the entire Redstone framework.

FSR Redstone is the comprehensive name for AMD’s full suite of ML-powered technologies, which includes:

  • Upscaling
  • Frame Generation
  • Ray Regeneration
  • Radiance Caching

The complete FSR Redstone suite remains exclusive to RDNA 4 hardware. However, AMD has noted that users on older cards can still utilize non-ML-based Frame Generation technologies. With the confirmed arrival of FSR 4.1 in July for RX 7000 users, the gap between AMD’s latest innovations and its previous generations continues to narrow.