If you've been a PC gamer for the past five years, then you've no doubt heard the term Frame Generation. It's at the forefront of marketing for both AMD's and Nvidia's best graphics cards and has slowly become a more important defining factor of performance, eclipsing that of even native figures as games have become more demanding.
But what is Frame Generation? More importantly, how does it work? That's what TechRadar is here to answer. Making things a little more confusing is the fact that the two companies have different names and versions for their respective technologies. We're going to cut through all the jargon and get you up to speed.
Before we go any further, however, it's important to note that you'll need a fairly modern graphics card to use Frame Generation, regardless of the company making it. For AMD, that means a minimum of an RX 5000 GPU (though later software is exclusive to the latest GPUs), and only Nvidia's RTX 40 series and RTX 50 series from 2022 and 2025 can access the technology.
What do frames mean?
In computing terms, frames are images generated by your graphics card which are sequenced for a framerate. This will mean a minimum of 30 frames (but ideally 60 or above) to make games playable. Frame Generation adds an AI-generated frame alongside a natively rendered one, and Nvidia's Multi Frame Generation can add up to three alongside the natively rendered frame.
What is AMD Frame Generation?
Frame Generation is a tool for your graphics card that generates additional frames when gaming. This allows for much higher framerates compared to the native performance of the GPUs, and has essentially become a requirement in recent years for acheiving high framerates in higher resolutions in demanding software.
With that said, the technology works differently on AMD and Nvidia hardware. While AMD originally launched its Frame Generation technology with AMD Fluid Motion Frames (AFMF), which is currently in its second iteration, the tech has some limitations. While open source and driver-based to run through software, it provided a way for older (or slower) graphics cards to get a framerate boost but was superseded by AMD FidelityFX Super Resolution (FSR). Though FSR 3 (available to RX 5000 series cards and up) and FSR 4 (RX 9000 only) can do a lot more.
FSR 3 was launched in September 2023 with more advanced algorithmically-based upscaling and Frame Generation technology only for supported games, rather than being driver-based and open-source. With FSR 3's upscaling and Frame Generation enabled, you could see a massive uplift in performance compared to native rendering, especially in higher resolutions, sometimes as much as 3x.
In contrast, the recently released FSR 4 is AI-powered and exclusive to the RX 9000 series with its Frame Generation technology. That means only the RX 9070 and RX 9070 XT can make use of the quality improvements and upgradable API, which allow for superior image quality and smoothness when compared to the previous version. It uses the same FSR 3.1 API and now features "anti-lag" features to remove input latency, which is a common concern with interlaced frames.
What is DLSS Frame Generation?
Things are more straightforward with Nvidia's Frame Generation technology, which has seen a similar evolution since it was first introduced. It begins with Nvidia's Deep Learning Super Sampling (DLSS) which is an AI-powered upscaler utilizing the Tensor Cores on RTX graphics cards. By using hand-tuned algorithms, the native rendering is halfed by resolution and then upscaled with AI to achieve higher framerates compared to what could be done natively.
DLSS 3 launched in October 2022 with the release of Nvidia's RTX 40 series graphics cards and the exclusive feature of Frame Generation, something that previous RTX cards could not do. This is due to the requirement of fourth-generation Tensor Cores which were fast enough to generate frames. Described by the company as the "performance multiplier", Nvidia's Frame Generation uses a combination of software and hardware for up to 4x the native performance and twice the responsiveness.
On the software side, there's AI Super Resolution, AI Frame Generation, and Nvidia Reflex to upscale the picture, generate additional frames with AI, and then reduce input latency. This is done on the hardware front with 1.4 PetaFLOPs available in the fourth-generation Tensor Cores, 305 TOPs of the Optical Flow Accelerator and Nvidia's supercomputer, which is always running and improving DLSS algorithms.
That brings us to DLSS 4, released on January 30, 2025, exclusively for the RTX 50 series (such as the RTX 5090, RTX 5080, and RTX 5070 Ti), and evolved the technology with Multi Frame Generation (MFG). While the previous version generated a frame for every natively rendered one, MFG can generate three, allowing for 4x the native framerates. Nvidia claims that the latest AI model is 40% faster while using 30% less VRAM.
DLSS FG vs AMD FG
Now that we've outlined how Frame Generation works with AMD's and Nvidia's respective takes on the tech, we can look at the differences in performance and the supported games. AMD's FSR 4 currently supports nearly 40 games and counting, including Call of Duty: Black Ops 6, FragPunk, Civilization 7, and Marvel's Spider-Man 2. In contrast, there are around 76 games that support FSR 3, with some exclusive to FSR 3.1.
In contrast, there are currently over 150 games that support Nvidia's DLSS 3 and counting, with 75 games confirmed to support DLSS 4 soon. This includes support for games such as Alan Wake 2, Dragon Age: The Veilguard, Microsoft Flight Simulator 2024, Silent Hill 2, Indiana Jones and the Great Circle, and Marvel Rivals.
What you'll notice is that Frame Generation is not available for everything you could ever want to play on PC, whether it's handled by AMD or Nvidia hardware. While the open-source driver-based AMD AFMF can fill in the gaps by unsupported games, it will not deliver the same level of performance boost by comparison. Given the two most recent versions, it makes the most sense to compare FSR 4 vs DLSS 4 for the best possible performance.
While Nvidia DLSS 3's Frame Generation and DLSS 4's Multi Frame Generation have left a strong impression as the leading options, the gap is certainly closing as AMD's FSR 4 looks to hold its own. This is likely due to being exclusive to the RX 9000 series instead of being more open as with AFMF, FSR 2, and FSR 3 from before. We can see this is evident in Digital Foundry's benchmarking especially.
AMD's RX 9000 lineup is roughly equivalent to midrange RTX 50 series hardware, so the comparisons in upscaling performance with the RX 9070 XT and RTX 5070 Ti are about as even as things can get. While the general frame rates are a little lower in FSR 4 compared to FSR 3, it's generally said that the image quality is higher as a result, but you'll have the option of both on a recent AMD GPU.
We can see this with Ratchet & Clank: Rift Apart in 4K. The RX 9070 XT delivers average framerates of 109, where the RTX 5070 Ti averages 106. However, with FSR 4 Performance mode, that drops down to 101, whereas the RX 5070 Ti shoots up to 128. Things are comparable in terms of image and smoothness, but Nvidia pulls ahead slightly, albeit not quite as dramatically as you may be expecting.
While it's still early days for the latest versions of the Frame Generation tech, looking back to FSR 3.1 vs DLSS 3.7 can give a wider indication of how it performs for the previous-generation RX 7000 and RTX 40 series cards. Remember, both FSR 3 and DLSS 3 also work on the latest GPU generations, just not the other way around. We can see the quality and performance differences from Ultrawide Tech's benchmarking from late last year.
We can see what FSR 3's Frame Generation and DLSS 3's Frame Generation can do on the previous-generation graphics cards, AMD's 7900 XT and Nvidia's RTX 4070 Ti. Natively, neither card can achieve 60fps in Star Wars Jedi: Survivor in 1440p on the Epic preset, but upscaling (FSR and DLSS) gets boosted to 76fps and 60fps, respectively. Frame Generation, however, greatly improves things to 147fps and 98fps apiece. This also makes 4K60 not only possible but also goes beyond these figures, as the former gets close to 120fps and the latter excels to nearly 80fps.
Demanding games, such as Cyberpunk 2077, are a tall order for native hardware in higher resolutions than 1080p. The RX 7900 XT and 4070 Ti can barely achieve 30fps in 1440p with the game at Ultra preset with ray tracing enabled. Through Frame Generation, the figures climb to 94fps and 85fps each, with both cards able to achieve 4K60 (and above).
Is Frame Generation good?
While we've been consistently impressed with what AMD and Nvidia can do with Frame Generation (particularly with the recent versions), the tech is not always going to be a win-win solution. Latency is a big concern of Frame Generation even with AMD using an "anti-lag" feature with FSR 4 and Nvidia Reflex coming in clutch for DLSS 3 and DLSS 4. As such, with AI generated frames interpolated between rendered ones, it's not going to be the ideal way to play certain games that require immediate inputs. You wouldn't want to have lags and delays in competitive games like CS2, Marvel Rivals, and Rainbow Six: Siege, even though the tech is no doubt impressive.
That's the trade-off with Frame Generation as a concept. However minor, you're going to notice a slight delay with the potential for fuzziness, visual artifacts, and motion blur compared to native rendering. It was a big problem when this technology was new, however, it's been improved year-on-year with each new version from AMD and Nvidia to a point where it's much less of a problem than it once was. While the tech has been seen as black magic by some, it does have its limitations when it comes to image quality and smoothness.
Overall, Frame Generation is an excellent way to improve framerates on AMD and Nvidia graphics cards if you can look beyond the minor caveats. FSR 4's quality mode and DLSS 4's Multi Frame Generation offer sweeping improvements on the previous versions that make them as close to native-looking as possible with framerates you just won't achieve conventionally.
The pushback on Frame Generation relates to the debate around "fake frames". It's been a point of discussion for over three years now, with some PC gamers decrying the tech as the death of native performance. It's the distinction between natively rendered and AI-generated frames, after all. We're seeing more and more titles rely on this technology to hit playable framerates, which, consequently, has seen the optimization of some PC games suffer as a result. Some have seen Frame Generation as a crutch, which it can be if a title runs poorly by default, but that's too sweeping of a blanket statement to make for emerging technology.
We have to accept that games are becoming more advanced, requiring more VRAM and more data to install and run. That's to say nothing of ray tracing, which is incredibly demanding on the hardware even in 2025; some games are even being released now with ray tracing on by default, meaning you have to use AI upscaling and Frame Generation if you want playable figures. That's the current landscape for PC gaming, and it's unlikely to change any time soon.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
