Jensen Huang isn't known for mincing words, and the Nvidia CEO didn't hold back when responding to the firestorm over DLSS 5. At a recent event, he told critics they're "completely wrong" about the technology - a bold stance when your core customer base is revolting.
The backlash centers on how much AI generation DLSS 5 uses versus traditional rendering. The new version can generate up to 31 out of every 32 frames using AI, with the GPU only rendering one actual frame. Gamers argue this crosses a line from "enhancement" to basically watching an AI hallucinate what the game should look like.
Huang's counterargument? That traditional rendering is already "fake" in many ways. Anti-aliasing, post-processing effects, even texture filtering - all of these are approximations and optimizations. DLSS 5 is just the next evolution. And from a pure technical standpoint, he's not entirely wrong. The output does look better in many cases.
But here's what the spec sheets don't capture: when only 3% of frames are actually rendered, are you playing a game or watching an AI's interpretation of a game? It's a philosophical question dressed up as a technical one. The technology is genuinely impressive - Nvidia's frame generation models are remarkably good at predicting motion and maintaining visual coherence. The question is whether players want it.
Some developers are already expressing concerns about debugging and optimization when the vast majority of what players see never actually came from their rendering pipeline. How do you fix visual glitches in AI-generated frames? How do you art direct something the AI is making up?
Huang may be technically correct that DLSS 5 produces good results. But telling your customers they're "completely wrong" for caring about how those results are achieved is a bold strategy. Gamers tend to have strong opinions about authenticity - and they're not easily convinced by benchmarks alone.
