A review of TechPowerUp and Tom’s Hardware US tests, which come to the same conclusion: the FSR 2.0 is a real success.
Technology FidelityFX Super Resolution 2.0 d’AMD is available for game Deathloop since May 12. During the presentation last March, AMD promised better image quality than the FSR 1.0, at the cost of slightly higher resource consumption. Verification through two tests; one made by W1zzard for the TechPowerUp site, the other by Aaron Klotz for the Tom’s Hardware US site. We will focus mainly on the second which has the advantage of precisely informing the average number of frames per second conferred by the FSR 1.0, FSR 2.0 as well as the DLSS of NVIDIA and this, for different graphics cards.
For the purely visual aspect, TechPowerUp offers the video below opposing the FSR 1.0, the FSR 2.0 and the DLSS 2.3 in several scenes of the game Deathloop. You can view full screen still images in different definitions on the source.
Some additional comparisons offered by Tom’s Hardware US in 1080p.
Les gains d’IPS
Regarding the gains in frames per second, AMD had announced the color, the FSR 2.0 is less saving than the FSR 1.0. At 1080p/Ultra, a Radeon RX 6700 XT averages 110 frames per second. The FSR 1.0 quality mode pushes the counter to 149.4, and 167.6 in performance mode. Under the same conditions and with the same modes, the FSR 2.0 tops out at 135.8 IPS in quality mode, 152.2 IPS in performance mode. In any case, the gains compared to native rendering remain of course significant.
With a GeForce RTX 3070, the performance differences are very small between DLSS, FSR 1.0 and FSR 2.0 at 1080p. These technologies provide about 10 images per additional compared to the native definition.
Same experience but in 2160p this time. While FSR 1.0 quality mode allows the Radeon RX 6700 XT to average above 60 FPS, FSR 2.0 with the same setting lowers the framerate at around 50 fps. In the case of the GeForce RTX 3070, DLSS Performance offers the biggest boost in FPS.
| Script | Radeon RX 6700 XT | GeForce RTX 3070 |
| 2160p native | 35,1 | 35,9 |
| FSR 1.0 Quality | 60,8 | 57,7 |
| FSR 1.0 Performance | 79,7 | 74 |
| FSR 2.0 Quality | 51,4 | 52,3 |
| FSR 2.0 Performance | 67,5 | 67,1 |
| DLSS Quality | – | 52,4 |
| DLSS Performance | – | 76,9 |
What about older graphics cards?
Since not everyone is lucky enough to have a graphics card as recent as a Radeon RX 6700 XT or GeForce RTX 3070our colleague also made some measurements with older references: GeForce GTX 1080, GeForce GTX 970, RX Vega 64, RX 480, in 1080p / Ultra for all, except the GTX 970 proven in 1080p / High.
| Script | GeForce GTX 1080 | GeForce GTX 970 | RX Vega 64 | RX 480 |
| 1080p native | 73 | 39 | 68,5 | 37,1 |
| FSR 1.0 Quality | 93,2 | 48,4 | 86,3 | 44 |
| FSR 1.0 Performance | 107,4 | 53,8 | 98,6 | 48,3 |
| FSR 2.0 Quality | 86,1 | 45,1 | 80,3 | 40,2 |
| FSR 2.0 Performance | 96,9 | 48,5 | 90,1 | 44,2 |
Conclusion
In conclusion, Aaron Klotz believes that AMD “did a great job with FSR 2.0 finding a way to deliver image quality similar to DLSS 2.0, without the hardware requirements that NVIDIA’s upscaling technology requires” ; the FSR 2.0 actually has the advantage of accommodating a wide range of GPUs, including old and competing ones.
Same story for W1zzard from TechPowerup. The latter writes: “AMD has done the unthinkable: the new FidelityFX Super Resolution FSR 2.0 is excellent, as good as DLSS 2.0, even DLSS 2.3. Sometimes it does a little better, sometimes a little worse, but overall it’s a huge win for AMD.”.
At this stage, the only question finally concerns the adoption of FSR 2.0, subject to the goodwill of the developers who must implement the technology in their productions. For now, the only game to benefit from FSR 2.0 is Deathloop. AMD has formalized a dozen titles that will soon adopt it. For comparison, in just under a year, the FSR 1.0 had gone from 7 games at launch to over 80.
Sources : TechPowerUP, Tom’s Hardware US