Introduction
I will be doing an in-depth comparison between AMD’s FidelityFX Super Resolution (FSR) 2.0 and Nvidia’s Deep Learning Super Sampling (DLSS). Both FSR 2.0 and DLSS are upscaling technologies that allow games to render at lower resolutions, then use advanced upscaling techniques to increase the resolution while maintaining image quality. This improves game performance and allows higher graphics settings to be used.
How FSR 2.0 Works
AMD’s FSR 2.0 uses temporal upscaling combined with a spatial upscaler.
-
It first uses data from previous frames to estimate the movement of each pixel. This is known as motion vectors.
-
It then uses this motion vector data to upscale and reconstruct a higher resolution image from the lower resolution input.
-
An edge-directed spatial upscaler is then applied to further enhance fine details.
-
This two-pass approach produces a very sharp final 4K image that is close to native resolution, but from a lower 1440p or 1080p source.
FSR 2.0 Advantages
-
The main advantage of FSR 2.0 is that it is an open source technology that works across multiple GPUs and platforms.
-
It can run on Nvidia, AMD, and even integrated graphics from Intel and AMD. This makes it widely accessible.
-
It also has a very low performance hit of just 2-3% when enabled.
How DLSS Works
Nvidia’s DLSS utilizes deep learning and AI to upscale games.
-
It uses a special Tensor Core AI accelerator that is present on RTX GPUs to run a deep neural network.
-
This deep learning model has been trained on millions of high resolution images to intelligently upscale and reconstruct detail from lower resolution inputs.
-
The AI analyzes each individual frame and intelligently applies directional details, textures, and edges to construct a realistic higher resolution image.
DLSS Advantages
-
DLSS produces extremely sharp 4K images that can rival native resolution in quality.
-
It also gives a significant performance boost versus native resolution rendering, with up to 2-3x higher FPS in some games.
-
DLSS has wide support and has been integrated into over 200 games. The extensive training of the AI model allows it to work well in almost any game scenario and art style.
Image Quality Comparison
In terms of image quality, both FSR 2.0 and DLSS can produce very good results, with the 4K upscaled image looking close to native 4K quality.
However, Nvidia DLSS generally produces somewhat sharper and more detailed images compared to FSR 2.0. This is because the deep learning model has been extensively trained on high resolution game imagery.
AMD FSR 2.0 relies on more generalized upscaling algorithms rather than game-specific training data. So some fine details can get lost or smoothed over in the upscaling process with FSR 2.0.
But AMD FSR 2.0 still delivers excellent image quality that is extremely close to native resolution, especially when using the Ultra Quality mode. The image below demonstrates this on a sample game screenshot:
**FSR 2.0 vs DLSS image quality comparison**
*Credit: RockPaperShotgun*
As we can see, FSR 2.0 Ultra Quality mode and DLSS Quality mode produce nearly equivalent results to native 4K, while giving a massive performance boost.
Performance Impact
Both technologies offer significant performance gains over native resolution rendering.
DLSS generally provides a higher performance boost, with up to 3x higher FPS reported in some games. With FSR 2.0, a 1.5-2x FPS gain is more typical.
The chart below compares FSR 2.0 and DLSS performance in Horizon Zero Dawn at 4K resolution using an RTX 3080 Ti:
| Rendering Technique | Average FPS | Performance Gain |
|-|-|-|
|Native 4K| 87 fps| 1x |
|FSR 2.0 Quality|143 fps| 1.64x |
|FSR 2.0 Balanced|157 fps| 1.80x |
|FSR 2.0 Performance|174 fps| 2.00x |
|DLSS Quality|155 fps| 1.78x |
|DLSS Balanced|185 fps| 2.12x |
|DLSS Performance|203 fps| 2.33x |
As we can see, DLSS provides up to a 2.33x FPS gain, while FSR 2.0 reaches up to 2.0x. Both provide excellent performance improvements, but DLSS has an advantage.
Supported GPUs and Games
One major advantage of FSR 2.0 is its wide hardware support. FSR 2.0 works on Nvidia, AMD, and Intel GPUs. Even budget and integrated graphics can benefit.
DLSS only works on Nvidia RTX cards as it requires the specialized Tensor Core AI hardware. Older Nvidia GTX cards cannot use DLSS.
Both technologies have extensive game support. Over 200 games now support DLSS, while over 50 games support FSR 2.0. Since FSR 2.0 is open source, it is easier for developers to add support.
Conclusion
Overall, both AMD FSR 2.0 and Nvidia DLSS are excellent solutions for boosting game performance while maintaining high image quality.
DLSS currently provides superior image quality and higher performance gains, but requires an Nvidia RTX GPU.
FSR 2.0 offers excellent visual results and performance gains on any GPU, but can sometimes have slightly softer image quality compared to DLSS.
I recommend enabling these upscaling technologies wherever available. They provide a great experience boost and allow higher resolutions and settings to be used. DLSS has a slight edge right now, but AMD FSR 2.0 is an impressive technology as well, especially considering its open source nature and wide compatibility. Both are leaps ahead of traditional upscaling methods.
