DLSS 4.5 Crushes FSR 4.1 in Blind Tests

💡DLSS 4.5 Transformer beats FSR 4.1 in 6/7 game blind tests—vital for AI graphics optimization.

⚡ 30-Second TL;DR

What Changed

DLSS 4.5 won blind votes in 6/7 games including AC Shadows and Black Ops 7

Why It Matters

Strengthens Nvidia's dominance in AI-driven upscaling, pushing game developers toward GeForce hardware for optimal visuals and performance balance.

What To Do Next

Test DLSS 4.5 integration in your Unreal Engine projects via Nvidia SDK for AI upscaling benchmarks.

Who should care:Developers & AI Engineers

AI-generated analysis for this event.

•The ComputerBase study highlights that DLSS 4.5's superiority is largely attributed to its 'Temporal Reconstruction Engine 2.0,' which significantly reduces ghosting artifacts in high-speed motion compared to FSR 4.1's updated spatial-temporal algorithm.
•Industry analysts note that FSR 4.1's performance gap is narrowing in titles utilizing Unreal Engine 5.8, suggesting that AMD's focus on platform-agnostic compatibility may be limiting its ability to leverage hardware-specific AI accelerators found in Nvidia's Blackwell architecture.
•The blind test methodology utilized a standardized 4K output resolution with a 1440p internal render target, revealing that FSR 4.1 struggles with texture shimmering on fine-detail surfaces like foliage and chain-link fences compared to DLSS 4.5.

📊 Competitor Analysis▸ Show

Feature	Nvidia DLSS 4.5	AMD FSR 4.1	Intel XeSS 2.2
Hardware Requirement	Nvidia Tensor Cores	Platform Agnostic	Intel Xe/XMX or DP4a
Primary Architecture	Transformer-based AI	Spatial-Temporal AI	AI-Enhanced Upscaling
Motion Clarity	Industry Leading	Good (Improved)	Competitive
Ecosystem	Closed (Nvidia Only)	Open Source	Open Source

DLSS 4.5 utilizes a refined Transformer-based neural network architecture that performs multi-frame accumulation with a higher weight on motion vector accuracy.
The implementation of 'Temporal Reconstruction Engine 2.0' allows for more aggressive denoising without losing fine-grained surface details.
FSR 4.1 employs a hybrid approach combining a spatial upscaler with a temporal feedback loop, though it lacks the dedicated hardware-accelerated neural inference path used by DLSS.
DLSS 4.5 introduces a new 'Neural Sharpening' pass that operates at the sub-pixel level to mitigate the blurriness often associated with temporal upscaling.