Jensen Huang Says DLSS 5 Critics Have It “Completely Wrong”
Nvidia CEO Jensen Huang is pushing back — hard — on critics who have characterised DLSS 5 (Deep Learning Super Sampling 5) as just another form of generative AI upscaling. His position: they’ve misunderstood the technology entirely.
The debate matters. DLSS 5 is not an academic argument. For the majority of PC gamers in Southeast Asia running mid-range RTX hardware — RTX 3060, RTX 4060, and their equivalents — it’s one of the primary tools for getting playable frame rates in modern titles without upgrading.
What Huang Actually Said
Speaking in response to a wave of criticism from PC hardware commentators and developers, Huang argued that DLSS 5 is “very different than generative AI” — and that critics who have framed it as AI-generated imagery filling in frames are, in his words, “completely wrong.”
The distinction Huang is drawing is technical and specific. Generative AI — in the way most people understand it — synthesises new content from a model trained on large datasets, producing output that wasn’t derived directly from existing input data. DLSS 5, Nvidia argues, uses neural rendering informed by game engine data: motion vectors, depth buffers, and prior frame information that the engine itself provides. The network reconstructs and upscales using real spatial and temporal data from the scene, not by generating plausible-looking imagery from a generalised model.
Critics have pushed back on that framing, arguing that the line between “reconstruction informed by a neural network” and “generative AI” is blurry at best — particularly with DLSS 5’s multi-frame generation capabilities, which can produce multiple output frames from a single rendered frame. That’s a meaningful leap from the temporal accumulation approach DLSS used in its earlier versions.
Huang has not conceded the point. Nvidia’s position remains that the deterministic, engine-fed nature of the pipeline places it in a fundamentally different category.
Why This Debate Lands Differently in SEA
The framing war between Nvidia and its critics is interesting. What matters more to most PC gamers in the region is the second-order question: does the output hold up in competitive conditions?
For a Valorant or CS2 player in Malaysia or Indonesia running a mid-range GPU, DLSS 5’s frame generation is not a luxury feature — it’s the difference between a stable 144fps and a fluctuating 80. Competitive play at that frame rate range has a measurable impact on reaction time windows. A monitor running at 165Hz paired with DLSS 5’s multi-frame output is a genuinely different experience from one without it, regardless of what you call the underlying process.
The “is it really AI?” debate is, to some extent, a concern for enthusiasts benchmarking at 4K on flagship hardware. At the RTX 4060 tier — which dominates sales across SEA storefronts — the practical argument for DLSS 5 is strong. Latency introduced by frame generation remains a legitimate concern for competitive players, and Nvidia’s Reflex integration is the counter-argument Nvidia ships alongside it. Whether that trade-off is acceptable depends on the game, the title’s netcode, and the player’s sensitivity to input lag — not on how the frames are categorised philosophically.
Huang’s insistence that critics have the technology “completely wrong” may or may not change the debate. The benchmarks will.
Source: Rock Paper Shotgun
Marcus Tan | Esports Analyst | Based in Singapore
