Wednesday, December 4, 2024
Hardware Guide

DLSS & FSR (Beginner Explanation) – Which Games & How does it work?

Unless you’ve been living under a rock, you probably already know about DLSS and FSR. This post isn’t here to tell you that they can boost your frame rates—you likely already know that. Instead, it’s meant to explain the differences between the two technologies and provide a simple explanation of how they work, followed by a slightly more technical breakdown.

Finally, we’ll cover the games that support them.

Simple Explanation: DLSS

 Baking a Cake with AI

Imagine you’re baking a cake, but you’re short on ingredients (low-resolution rendering). You hand over the job to a genius chef (AI), who not only makes the cake but also adds extra layers and decorations based on their training with thousands of cakes they’ve studied (AI training data).

  • The chef doesn’t just stretch out the cake to look bigger—they rebuild it with high-quality layers and icing so it tastes (or looks) just like a full-sized cake.
  • They even preemptively add some finishing touches (frame generation) to make it look smoother and more appealing.

Result: A cake that looks and feels as good as a full-sized one, made with the help of an intelligent assistant.

Simple Explanation: FSR

Decorating a Pre-Baked Cake

Now imagine you’re handed a simple, smaller cake (low-resolution render), and your job is to make it look bigger and more appealing. Instead of rebuilding it from scratch, you add icing, decorations, and sprinkles to make it appear more polished.

  • The process is faster and works with any kind of cake  but it’s not as refined as having a master chef rebuild it from the ground up.
  • The closer someone looks, the more they’ll notice the shortcuts, like uneven layers or smudged decorations.

Result: A cake that looks good from a distance and satisfies most needs, but it doesn’t have the intricate details of the DLSS cake.

The left side shows a smaller polished and detailed cake while the right side shows a BIGGER but less polished cake. This depicts the difference: FSR generates more framerates (with lower quality), DLSS less framerates (with higher quality).

Technical Explanation: DLSS

  • Step 1: Lower-Resolution Rendering
    Your GPU renders the game at a lower resolution (e.g., 1080p instead of 4K). This saves performance because lower-res images require less work for the GPU.
  • Step 2: AI Upscaling
    NVIDIA’s AI model, trained on high-quality image data, analyzes the low-res frame and predicts what a higher-resolution version would look like. It fills in missing details using advanced algorithms powered by Tensor Cores (specialized AI hardware on RTX GPUs).
  • Step 3: Frame Refinement
    DLSS smooths out edges, enhances textures, and creates a final frame that looks like native high resolution (or even better).
    With DLSS 3, it can also generate entirely new frames between existing ones, boosting frame rates significantly.

 

Technical Explanation: FSR

FSR (FidelityFX Super Resolution)

  • Step 1: Lower-Resolution RenderingJust like DLSS, FSR starts by rendering the game at a lower resolution (e.g., 1080p instead of 4K). This saves performance for the GPU.
  • Step 2: Upscaling and SharpeningFSR doesn’t use AI. Instead, it uses a spatial or temporal upscaling technique.
    • Spatial Upscaling (FSR 1.0): Enhances each frame individually by sharpening edges and textures.
    • Temporal Upscaling (FSR 2.0): Uses data from multiple frames (motion and depth info) to create smoother and more detailed upscaled frames.
  • Step 3: Final FrameFSR produces a high-resolution image but relies more on sharpening and contrast adjustments to make the image look detailed, which can sometimes introduce minor artifacts or blurriness.

DLSS & FSR – Benchmark Examples

Below Im running Hogwarts Legacy with a 3050RTX (Desktop). At ultra 1440p you can expect  20-25 extra framerates with DLSS.

DLSS vs FSR – StarField Medium (1440p)

Like we discussed above. FSR will give you more frameartes than DLSS but the quality is going to be lower. In the example above, Starfield at medium (1440p) runs at 37FPS on avg. without FSR/DLSS. Then you get an extra 11fps with DLSS and 29FPS with FSR (~30% & ~100% respectively).


List of Games with FSR & DLSS Support

Game Title DLSS Support DLSS Version FSR Support FSR Version
Counter-Strike 2 (CS2) No N/A No N/A
Fortnite Yes 3.0 Yes 2.0
Grand Theft Auto V (GTA V) No N/A No N/A
League of Legends (LoL) No N/A No N/A
Minecraft Yes 2.0 No N/A
Call of Duty: Warzone Yes 2.0 Yes 1.0
World of Warcraft No N/A Yes 1.0
Overwatch 2 No N/A No N/A
Valorant No N/A No N/A
Elden Ring No N/A No N/A
Cyberpunk 2077 Yes 3.5 Yes 2.1
Red Dead Redemption 2 Yes 2.0 Yes 2.0
The Witcher 3: Wild Hunt Yes 3.0 Yes 2.1
God of War Yes 2.0 Yes 2.0
Starfield No N/A Yes 2.0
Baldur’s Gate 3 No N/A Yes 2.0
Resident Evil 4 No N/A No N/A
Hogwarts Legacy Yes 3.0 Yes 2.0
Apex Legends No N/A No N/A
Spider-Man: Miles Morales Yes 3.0 Yes 2.1
Spider-Man Remastered Yes 3.0 Yes 2.1
Dota 2 No N/A Yes 1.0
The Last of Us Part I Yes 2.0 Yes 2.0
Assassin’s Creed Valhalla No N/A Yes 1.0
Halo Infinite No N/A No N/A
Far Cry 6 Yes 2.0 Yes 1.0
Watch Dogs 2 No N/A No N/A
PUBG: Battlegrounds No N/A No N/A
Horizon Zero Dawn No N/A Yes 1.0
Ghostwire: Tokyo Yes 2.0 Yes 2.1
Death Stranding Yes 2.0 Yes 2.0
Dying Light 2 Yes 3.0 Yes 2.0
Hitman 3 Yes 2.0 Yes 2.1
Marvel’s Guardians of the Galaxy Yes 2.0 Yes 2.0
Forza Horizon 5 No N/A Yes 2.2
Uncharted: Legacy of Thieves Collection Yes 2.0 Yes 2.1
Returnal Yes 3.0 Yes 2.0
Atomic Heart Yes 3.0 Yes 2.0
Days Gone No N/A Yes 2.0
Deathloop Yes 2.0 Yes 2.0
Dead Island 2 No N/A No N/A
Shadow Warrior 3 No N/A Yes 1.0
Alan Wake 2 Yes 3.0 No N/A
Remnant II Yes 3.0 No N/A
A Plague Tale: Requiem Yes 2.0 No N/A
Palworld No N/A No N/A
Evil West Yes 2.0 No N/A
The Lord of the Rings: Gollum Yes 3.0 No N/A
Forspoken Yes 2.0 Yes 2.0
Atlas Fallen No N/A No N/A
Star Wars Jedi: Survivor No N/A No N/A
Sons of the Forest No N/A No N/A

Note: The availability of DLSS and FSR support, as well as their versions, can change with game updates. It’s advisable to check the latest patch notes or official game websites for the most current information.

DLSS and FSR Versions

Both DLSS (by NVIDIA) and FSR (by AMD) have evolved over time, introducing new features and improvements in each version. Here’s a simple guide to understand what each version offers.

DLSS 1.0

  • The Basics: Introduced in 2019, it was NVIDIA’s first attempt at AI-based upscaling. It used AI to upscale lower-resolution images to higher resolutions.
  • Issues: Image quality was often blurry, and the performance gains weren’t consistent across all games.
  • Games: Very few supported it initially, such as Final Fantasy XV.

DLSS 2.0

  • Big Improvements: A complete overhaul. This version significantly improved image quality, making it look close to or even better than native resolution in some cases.
  • How It Works:
    • Uses AI models trained on general data, making it adaptable to many games.
    • Sharpens edges and textures for a more natural look.
  • Advantages:
    • Huge boost to frame rates.
    • Widely adopted in games like Cyberpunk 2077 and Control.

DLSS 3.0

  • Frame Generation Introduced: This is the big one. DLSS 3 doesn’t just upscale images; it creates entirely new frames using AI.
    • Example: If your GPU produces 60 FPS, DLSS 3 can generate enough frames to make it feel like 120 FPS.
  • Who Can Use It:
    • Exclusive to RTX 40-series GPUs because it relies on the Optical Flow Accelerator and advanced Tensor Cores.
  • Ideal Use: Great for single-player and visually demanding games like Cyberpunk 2077 with ray tracing enabled.

DLSS 3.5

  • Introduced Ray Reconstruction:
    • Uses AI to improve the quality of ray-traced effects (e.g., lighting, shadows, reflections).
    • Removes noise and enhances realism, especially in scenes with heavy ray tracing.
  • Works on More GPUs: Unlike DLSS 3.0, ray reconstruction works on RTX 20-series and 30-series GPUs, too.

FSR Versions

FSR 1.0

  • Basic Upscaling: Launched in 2021, it was AMD’s answer to DLSS but without AI. It upscales low-res frames and sharpens the image.
  • How It Works:
    • Spatial Upscaling: Processes each frame individually to enhance clarity.
  • Issues: Quality wasn’t great in fast-moving scenes, and artifacts were noticeable.
  • Compatibility: Worked on almost any GPU (AMD, NVIDIA, and Intel), making it very accessible.

FSR 2.0

  • Smarter Upscaling:
    • Introduced temporal data, meaning it uses information from multiple frames to improve the final image.
    • Reduced artifacts and improved quality, especially in motion-heavy scenes.
  • Improved Image Quality:
    • Comes closer to DLSS 2.0 in sharpness and detail.
  • Compatibility: Still works on most GPUs, including older ones.

FSR 3.0

  • Frame Generation Added: AMD’s response to DLSS 3.0. FSR 3 introduces frame generation, creating additional frames to improve smoothness.
  • Advantages:
    • Works on AMD RDNA 3 GPUs and later (Radeon 7000 series).
    • Open source, so developers can integrate it more broadly.
  • Challenges: Frame generation isn’t as widely supported yet compared to DLSS 3.

Comparison of DLSS/FSR Versions

Feature DLSS 2.0 DLSS 3.0 DLSS 3.5 FSR 1.0 FSR 2.0 FSR 3.0
Upscaling AI-based AI-based AI-based Spatial Temporal Temporal
Frame Generation No Yes (AI-generated) Yes No No Yes
Ray Tracing Boost No Yes Yes (Ray Reconstruction) No No No
GPU Compatibility RTX GPUs (20+) RTX 40-series only RTX GPUs (20+) All modern GPUs All modern GPUs RDNA 3 GPUs+
Best Use Case General Gaming Frame Rate Boost Ray-Traced Games Budget GPUs Smoother Quality Frame Rate Boost

Summary

DLSS (Deep Learning Super Sampling) by NVIDIA and FSR (FidelityFX Super Resolution) by AMD both enhance gaming performance and visuals, but they cater to different audiences:

  • DLSS:
    • Uses AI for superior image quality and smoother gameplay.
    • Exclusive to NVIDIA RTX GPUs (20-series and newer).
    • DLSS 3 adds frame generation, doubling FPS in supported games.
    • Best for high-end setups and ray-traced games.
  • FSR:
    • Hardware-agnostic, working on almost all GPUs (AMD, NVIDIA, and Intel).
    • Easier to implement and widely compatible.
    • Provides solid performance gains, though image quality slightly trails DLSS.
    • Ideal for budget and older systems.

So basically…

  • DLSS if you have an NVIDIA RTX GPU and prioritize cutting-edge features like AI-driven frame generation.
  • FSR if you’re on a budget, have a non-NVIDIA GPU, or need broader compatibility.

Author Profile

Miguel Salas
Miguel Salas
I am physicist and electrical engineer. My knowledge in computer software and hardware stems for my years spent doing research in optics and photonics devices and running simulations through various programming languages. My goal was to work for the quantum computing research team at IBM but Im now working with Astrophysical Simulations through Python. Most of the science related posts are written by me, the rest have different authors but I edited the final versions to fit the site's format.

Miguel Salas

I am physicist and electrical engineer. My knowledge in computer software and hardware stems for my years spent doing research in optics and photonics devices and running simulations through various programming languages. My goal was to work for the quantum computing research team at IBM but Im now working with Astrophysical Simulations through Python. Most of the science related posts are written by me, the rest have different authors but I edited the final versions to fit the site's format.

Leave a Reply

Your email address will not be published. Required fields are marked *