• 3 min read
DLSS 5 adds controls, but gamers remain unconvinced
Nvidia adds scene-level controls to DLSS 5, but gamers still question its visual impact, hardware demands, and limited customization.
Image: TechRadar
Nvidia used its SIGGRAPH 2026 keynote to reveal more about DLSS 5, the neural-rendering technology that drew a hostile response when it was first announced at GTC 2026. The update gives developers more control over how the system changes a game’s visuals, but it has not silenced the criticism.
According to Tom’s Hardware, Nvidia presented the technology in Los Angeles as a major part of its keynote. The company emphasized that DLSS 5 has changed since its initial reveal, when critics argued that its AI-generated enhancements made games look worse, particularly around faces, and could undermine a developer’s intended art direction.
How developers can control DLSS 5
Nvidia now offers developers three AI models with different levels of detail and performance impact. Developers can select among them on a scene-by-scene basis rather than applying one setting across an entire game.
Recommended reading
Natural raises $30 million for AI agent payments
They will also get:
- Two intensity sliders for adjusting the effects' “structure” and “tone.”
- Asset masks that prevent DLSS 5 from modifying selected objects or areas.
The masking system could allow developers to exclude characters' heads and faces while applying the technology to background objects and environmental details. They could also reduce the effect’s intensity for a subtler result.
DLSS 5 is not a replacement for Nvidia’s existing upscaling and frame-generation systems. Instead, it sits on top of that suite as an AI-based, photorealism-oriented filter driven by a new neural-rendering model.
Nvidia’s full presentation is available on YouTube, with the DLSS 5 section appearing near the beginning after the usual introduction.
Developers want more control and lower hardware demands
The new controls have produced a somewhat more positive response than the original announcement, but many gamers still consider them insufficient. One Reddit user described the technology as follows:
“I had so much hope for DLSS in the early days, what a dumpster fire it’s turned into.”
Another called it a “glorified Instagram filter.” The central complaint is that three models and two sliders do not provide the level of control developers may need to preserve a game’s visual identity.
DLSS 5 is expected to arrive in fall 2026, meaning sometime between September and November. Nvidia could still add more controls before release.
Performance is another unresolved issue. The original DLSS 5 demonstration ran on two RTX 5090 graphics cards, while Nvidia did not identify the hardware used for the SIGGRAPH presentation. The company did say the technology will run on a single graphics card and described it as “VRAM efficient,” without providing more specific requirements.
The source’s analysis expects the first release could require at least an RTX 5080, or possibly an RTX 5090, while also warning that early versions may suffer from artifacts. Nvidia’s original DLSS 1 had its own problems with blur and glitches before later versions became widely regarded as impressive upscalers. DLSS 5 still has time to improve, but Nvidia has not yet addressed the biggest concerns about control and performance.
AI Editor
Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.
via TechRadar
