Nvidia DLSS 4.5 dominates blind gaming test: 48% of players prefer it to AMD FSR and native

A new blind comparison of upscaling tech suggests GPUs today are judged as much on their AI tricks as on raw hardware. The findings matter because image reconstruction tools like Nvidia’s DLSS and AMD’s FSR now shape how games look and perform on millions of PCs.

What the blind test looked at

A recent blind study conducted by German site ComputerBase collected 6,747 votes from players across six well-optimized titles to see which rendering mode they preferred visually. Test participants were shown three side-by-side versions of the same scenes and asked to pick the best-looking image.

• DLSS 4.5 (set to Quality)

Nintendo Switch Presidents’ Day deals: 21 editor-picked games starting at $17

Steam Deck OLED scarcity hits stores as Valve warns: RAM shortfall could stall Steam Machine plans

• FSR 4 (set to Quality)
• Native rendering with TAA (Temporal Anti-Aliasing)

Key results at a glance

Out of the full sample, nearly half the votes went to one option. The headline figure: DLSS 4.5 captured roughly 48% of selections, giving it a clear lead over both AMD’s new upscaler and the native TAA output.

Metric	Value
Total votes	6,747
Top choice	DLSS 4.5 (~48% of votes)
Titles tested (highlights)	Satisfactory; Horizon Forbidden West; Cyberpunk 2077

Where the differences were largest—and smallest

In scenes from Satisfactory and Horizon Forbidden West, testers showed a strong preference for the Nvidia option; the gap between DLSS 4.5 and FSR 4 was pronounced in those high-detail environments. By contrast, the picture was more mixed in Cyberpunk 2077: a sizeable portion of voters—around one in four—saw little to no difference between Nvidia and AMD upscaling there.

That variability matters: it shows that game engine, lighting, scene complexity and resolution influence which upscaler performs best in practice.

Why DLSS often wins the eye test

One surprising result was that DLSS 4.5 frequently beat the native TAA output. The explanation lies in how each method reconstructs frames. TAA blends current and prior frames to reduce jagged edges, which can produce motion blur and softer details during movement. By contrast, DLSS replaces that pipeline with an AI-driven model trained on extensive image data to reconstruct sharper, more stable details in motion—particularly on fine geometry and textures.

FSR 4 has closed ground versus earlier versions, but the study noted remaining artifacts in some scenes: minor jagging on delicate geometry, ghosting around fast-moving objects, and shimmering in high-contrast highlights. Those imperfections contributed to lower preference scores in several test cases.

ComputerBase added a methodological caveat: because participants were asked to choose the single best image, the ranking doesn’t strictly reflect a linear “second/third best” perceived quality. The framing can skew how differences are reported in aggregate.

What this means for buyers and players

For consumers choosing a GPU today, the takeaway is practical: upscaling tech now meaningfully alters perceived image quality and should factor into hardware decisions—especially for players who prioritize visual fidelity at high resolutions.

• If you frequently play games with complex lighting and fine geometry, DLSS 4.5 appears to offer the most consistent visual uplift in the tested titles.
• FSR 4 shows solid improvement over older FSR versions, but may still show artifacts in motion-heavy or highly detailed scenes.
• Native TAA remains a viable baseline, but in many cases modern AI upscalers can deliver a crisper, more stable image.

Short-term, Nvidia retains an edge in perceived image quality in this blind comparison. Longer term, the gap may narrow as AMD refines its models and developers tune integrations per title. For now, the test underscores that image reconstruction algorithms are one of the most consequential—and visible—differences between GPU ecosystems.