A look at AMD's new FidelityFX Super Resolution, which aims to upscale games to higher resolution and competes with Nvidia's DLSS technology
Context & Ripple Effects
Nvidia set the template for ML-based game upscaling when it shipped DLSS 2.0 in 2020, claiming better performance and visuals from rendering at lower resolution — but locked behind RTX hardware. AMD's FidelityFX Super Resolution is its answer: an upscaler aimed at pushing games to higher resolutions without Nvidia's silicon requirement.
What made FSR consequential wasn't just the launch — its lineage spread widely. It later became the technical basis for Arm's Accuracy Super Resolution, a mobile upscaler released under an open-source license, and by 2025 AMD's own Radeon RX 9070 line was shipping with FSR 4 support, while Microsoft built a vendor-neutral DirectSR API around all three GPU makers' upscalers.
First-order effects
- Gamers on non-Nvidia hardware gain a working path to higher-resolution gaming, breaking DLSS's status as an RTX-exclusive feature and giving AMD a direct counter in the feature war.
- Nvidia no longer owns the upscaling narrative: every FSR adoption gives developers a reason to ship cross-vendor support rather than DLSS-only paths.
Second-order effects
- Intel validated the three-horse field when pre-release reviews of its XeSS upscaler on the Arc A770 showed competitive results against Nvidia's slightly older DLSS — a market AMD's move helped make necessary.
- Microsoft's DirectSR API emerged to let developers integrate Nvidia, AMD, and Intel upscaling through one interface, an abstraction layer that exists precisely because per-vendor upscalers fragmented development.
Third-order effects
- If the pattern holds, upscaling stops being a GPU differentiator and becomes commodity middleware — with differentiation migrating up-stack to frame generation (as in Nvidia's DLSS 4.5) rather than resolution reconstruction itself.
- Open licensing changes the battleground: FSR 2's reuse as Arm's mobile upscaler shows the technology competing on ubiquity across device classes rather than hardware lock-in.
The trend: GPU competition is shifting from raw silicon specs to AI upscaling ecosystems, where openness versus hardware lock-in decides how far each vendor's reach extends.