AMD Plans FSR 4 Upscaling for Laptops and Handhelds by End of 2026

AMD plans to bring its machine-learning-based FSR 4 to APUs, gaming laptops and handhelds by the end of 2026. The timeline came from Jack Huynh, senior vice president and general manager of Computing and Graphics, who oversees AMD's PC, graphics and semi-custom businesses. The Verge

Huynh disclosed the plan in comments to Korean publication The Elec. The same reporting was summarized as an expansion to APUs, gaming laptops and handhelds before 2027. Wccftech

According to that account, AMD will pursue the expansion in two directions. One track covers existing APUs. The other covers new product lines.

The Elec reported that the effort includes a new APU and a lightweight FSR 4 model for APUs, both expected by year's end. Huynh did not frame this as a research project. He described it as a shipping plan.

FSR 4 was first introduced for discrete external graphics cards, separate boards with their own power and cooling. AMD had previously promised FSR 4 for desktop graphics cards, not for APUs for laptops and handhelds. An APU combines CPU and graphics on one chip. AMD has said FSR 4 is exclusive to the AMD Radeon RX 9000 series. AMD

The constraint, as AMD executives have described it, is compute headroom, or spare capacity to run AI work. Executive David McAfee said AMD needs sufficient compute within a product for FSR 4 to make sense and wants to ensure the experience and quality is right.

A discrete card can budget power and chip area for ML inference, the live running of the model, that an APU, gaming laptop chip or handheld SoC cannot assume. The lightweight model is tuned to fit tighter power and thermal limits and to run alongside game workloads without eroding frame rate or adding latency.

Valve is already aligned with part of that roadmap. The company said it is working with AMD to bring FSR 4 to the Steam Machine's console-grade AMD APU. Valve has not committed to bringing FSR 4 to Steam Deck handhelds.

The broader context here is portability of ML-based reconstruction, where a game renders at lower resolution and AI rebuilds a sharp image. Backporting a discrete-GPU pipeline to integrated graphics is rarely a recompile. It forces choices about precision, model size, dispatch priority and where upscaling sits in the frame pipeline. Supporting existing APUs as well as new silicon implies two different engineering problems, one bounded by fixed hardware already in the field and one where hardware and model can be co-designed.

For developers and hardware planners following this, the distinction between existing and new APUs will be the detail to watch. For existing parts, compatibility lists and quality modes will define usefulness. For new parts, the question is how much inference throughput is reserved in the SoC and how consistently it is available under sustained load. In both cases, image stability at low internal resolutions will decide whether the feature earns regular use on small, power-limited devices.

In my view, the Valve data point is the most instructive early signal. A console-grade APU in a mains-powered enclosure is an easier first target than a battery-powered handheld. If FSR 4 lands cleanly there, then scales down to laptops and handhelds through the lightweight model, AMD will have threaded a path that preserves quality where compute is scarce. That would give integrated graphics a more credible route to higher effective resolution without higher native render cost.