What was tested
Biswas used an RTX 4070, a Ryzen 5 5600X and 16GB of RAM. He ran six games: Marvel Rivals, Control Resonant, Marvel's Spider-Man Remastered, Dragon's Dogma 2, Final Fantasy XVI and Monster Hunter Wilds. Every run used DLSS at the Quality preset. The variable was which frame generator sat on top.
Some games let you pick the upscaler and the frame generator separately. The article names Marvel Rivals and Spider-Man Remastered as examples. For the games that don't, he says he forced DLSS at the driver level and enabled FSR Frame Generation in-game. He doesn't say which titles needed that workaround or how he applied it.
The reported numbers
These are the author's figures, not independently verified measurements.
| Game | DLSS Quality only | + DLSS Frame Gen | + FSR Frame Gen |
|---|---|---|---|
| Marvel Rivals | 107 | 174 | 182 |
| Marvel's Spider-Man Remastered | 98 | 130 | 141 |
| Control Resonant | 89 | 101 | 97 |
| Dragon's Dogma 2 | 91 | 118 | 122 |
| Final Fantasy XVI | 71 | 93 | 85 |
| Monster Hunter Wilds | 87 | 105 | 112 |
Working out the gap between the two frame-generation columns gives:
- FSR ahead: Spider-Man Remastered by 11 FPS, Marvel Rivals by 8, Monster Hunter Wilds by 7 and Dragon's Dogma 2 by 4.
- DLSS ahead: Final Fantasy XVI by 8 FPS and Control Resonant by 4.
The headline claim is that FSR "outperforms" DLSS Frame Generation. That is a stretch for gaps of 4 to 11 FPS from what appear to be single runs. The article gives no repeat counts, no measurement tool and no resolution for the tests. It also doesn't give driver versions, game versions or the FSR version. The small gaps could easily sit inside run-to-run variation, and the article gives no way to tell.
Why this combination is possible at all
Mixing the two isn't a hack. It follows from a design change AMD made in FSR 3.1.
- In FSR 3.0, AMD's frame interpolator could only be used with its shader-based upscaler.
- With 3.1, AMD decoupled frame generation from the upscaling tech, so it can work with other upscalers such as DLSS or XeSS.
- TechPowerUp's review explains the mechanism. Earlier versions reused intermediate data from the upscaler step. FSR 3.1 adds an optional API so developers can generate that data separately, at a small GPU cost.
That last point matters. Developers have to wire it up, and the mix isn't guaranteed in every game. One Steam thread complained that The Last of Us Part I locks frame generation to FSR 3.1 upscaling or native AA, while Ratchet and Clank: Rift Apart allows DLSS or XeSS alongside it. A 2026 explainer from TechFuelHQ makes the same point. FSR 3.1 lets frame generation work with third-party upscalers when the integration provides the required motion vectors and depth data.
There is also a historical reason this comparison is now easy. Years ago, outlets noted that the decoupling would allow FSR 3 and DLSS 3 frame generation to be compared directly. Biswas has essentially done that on one machine.
The catch the FPS counter doesn't show: Reflex and latency
Biswas says using FSR Frame Generation on an Nvidia card costs him Nvidia Reflex. He says that matters more on keyboard and mouse than on a controller. That fits Nvidia's own description of the technology. When it launched DLSS 3, Nvidia said DLSS 3 integrations also incorporate NVIDIA Reflex, which synchronizes the GPU and CPU to keep system latency low. Nvidia's description is design intent, not a head-to-head test. It also doesn't confirm how Reflex behaves in every game when FSR is selected.
The key point is that displayed frames per second isn't the same as responsiveness. Generated frames raise the number on the counter. They don't make the game react to your mouse any faster. The article reports no latency measurements, so "better results" is only established for the frame counter.
AMD's guidance is to start from a healthy base frame rate, around 60 FPS, before turning on FSR 3 frame generation. All the baselines in this test sit above that, from 71 to 107 FPS. That is the situation where frame generation is most defensible. It says little about weaker systems.
A claim worth treating carefully
The article says upscalers and frame generation need "significant CPU bandwidth." It argues that stacking two vendors' technologies would be even more taxing, and that this is why developers bundle them. I couldn't verify that against primary sources. Nvidia's DLSS 3 announcement points the other way. Nvidia said DLSS Frame Generation runs as a post-process on the GPU, so it can raise frame rates even when the game is CPU-bound. That is the vendor's claim, and CPU-limited behavior varies by game. Still, the CPU-cost explanation is the author's assertion, not an established finding. Developers may bundle the features for simpler QA, compatibility or vendor partnerships. That is my reading, not a sourced statement.
The article's image-quality comments are also subjective. Biswas says he'd still pick DLSS over FSR and that FSR 3 still shows ghosting in vegetation. He offers no frame captures or blinded comparison, so treat that as one gamer's opinion.
Practical takeaways
If you have an RTX card and want to try this, a few cautions follow from the evidence.
- Check the game's menu first. If upscaling and frame generation are separate options and FSR Frame Generation is listed, you may not need any workaround.
- Change one variable at a time. TechFuelHQ's advice is to leave the resolution and upscaler alone, toggle only frame generation and replay the same scene. Then test a different upscaler in a separate pass.
- Judge feel, not only the counter. Compare mouse responsiveness with and without Reflex. Biswas's concern is specific to keyboard and mouse.
- Be wary of modifying files in online games. The article doesn't give exact driver settings, DLL versions or per-game steps. Multiplayer titles may object to file swaps or overrides, so check the game's own guidance first.
- Don't assume your results will match. This was one RTX 4070 with a mid-range CPU. Other GPUs, CPUs and resolutions could change the outcome.
The bottom line
The result is more modest than the headline suggests. FSR 3.1's decoupled frame generation lets an RTX owner mix vendors. On one test PC, that produced slightly higher reported frame rates in four of six games and lower ones in two. The cost is losing Reflex, and no latency data is offered to show whether that matters. For most players, the better question isn't which frame generator posts the bigger number. It's which one feels right in the games they play. Nvidia owners can now test that themselves in games that expose the separate options.
References
- I combined Nvidia DLSS with AMD FSR and got a result I didn't expect MakeUseOf · 2026-10-23T20:30:17+00:00
- AMD Announces FSR 3.1, Improves Super Resolution Quality, Allows Frame Generation to Work with Other Upscaling Tech techpowerup.com
- Frame Generation Explained: DLSS 4, FSR & AFMF (2026) techfuelhq.com