Nvidia launched its rather controversial DLSS 4 with its RTX 50 series, featuring better upscaling, ray reconstruction improvements, and a whole lot more. One feature that specifically made it so controversial was the introduction of multi frame generation, which can add up to three more frames to further boost the frame rate. DLSS was introduced in 2018 alongside RTX 20 series graphics cards, and while it's improved drastically and certainly has its uses, DLSS 4's multi frame generation works at its best when it doesn't really make sense to use it.

I'm not the first to make this point (nor will I be the last), but it's still a pretty big indictment of the technology. While its latency hit is pretty minimal, the times you'll use it are also the times where you're less likely to actually benefit from it.

Multi-frame generation is built for the high-refresh-rate monitors of tomorrow

And even then, it might not make sense

First and foremost, the latency hit ranges from minimal to severe depending on who you ask, but even if it's minimal, that doesn't mean you can just boost your 30 FPS to 120 FPS and have the same experience as someone playing natively at that frame rate. The game is still running at 30 FPS internally, which means it's only polling for inputs every 33.3ms. That's pretty noticeable, and regular DLSS 3.5 frame generation can feel jarring. DLSS 3.5 can insert a frame between two natively rendered frames, and in games like Spider-Man, it can feel awful.

As a result, you need a high base frame rate to avoid the feeling of jarring latency as you play. What that base frame rate needs to be will depend on what you're used to, but keep in mind that even at a base of 90 FPS, it can still feel strange on a 144 Hz monitor to play at 180 FPS, as your base FPS will actually diminish with multi frame generation enabled. This jarring feeling diminishes as base frame rates increase, with DLSS 3.5 frame generation being nearly unnoticeable once you start getting to around that 144 FPS marker.

Now, let's take that experience and apply it to DLSS 4. Instead of doubling your FPS, we're talking about potentially quadrupling it. The best experience you'll get when using it is when you're already getting a high frame rate, which in turn likely means that your base FPS is already capable of outputting to a 144 Hz monitor at the very least. And if you're already saturating your high refresh rate monitor, you'll get no benefit from any additional frames, as they don't decrease latency.

To make matters worse, Hardware Unboxed went so far as to say that Nvidia's marketing was "deceptive and misleading" when it comes to multi-frame generation. Latency in Alan Wake II goes from 31.8ms without any kind of frame generation all the way up to 45.4ms with 4x generation. That's a 43% increase in latency, and to get from 30 FPS to 120 FPS in the same game has an astonishingly high 108.4ms of latency.

On top of that, using multi-frame generation actually decreases the internal FPS of the game being played. Alan Wake II saw the internal FPS dip from 96 to 72 with 4x frame generation enabled according to Hardware Unboxed's testing, which meant that any massive FPS games were potentially negated by the decrease to "real" performance.

What this means is that to take advantage of 4x multi frame generation, you'll really need to have a very high-refresh rate monitor that can actually show you those frames. Render times aren't actually improving, so the additional FPS that you gain past the threshold of what your monitor supports won't result in any kind of improved gaming experience. And if you have a monitor that's 360 Hz or even 500 Hz, then chances are, you probably have some of the best PC hardware out there already to power it.

Multi frame generaton is a gimmick that requires a lot of thought to figure out

It's deceptive

Unlike other gimmicks that are pretty easy to identify as gimmicks, multi frame generation is in an interesting spot as it's a lot harder to actually identify it as a gimmick. More frames means a better gaming experience is the conventional wisdom that all PC gamers subscribe to, but it takes a significant amount of analysis to understand why it it may not be as good as it seems. This appears to have blinded many who may not be aware how much latency actually matters in a gaming sense.

To understand the problem, imagine playing a game at 30 FPS. It feels choppy and slow, right? At 30 FPS, multi frame generation may net you more than 100 FPS, but that same slow feeling will still exist, even if it's no longer choppy. This can feel pretty jarring, and the additional overheads introduced by frame generation itself could potentially make it feel worse than 30 FPS. As this scales with a higher base frame rate, the jarring feeling may diminish significantly, but then it becomes a matter of what your monitor is actually capable of displaying.

I don't blame gamers for not being aware of what multi frame generation is, but the actual benefits that it provides aren't necessarily what they seem once you look past the surface. It still makes the most sense to use it only when you're already getting a decently high frame rate, but then it also only makes sense to use it when you have a very high refresh rate monitor. Frames generated past the refresh rate of your monitor don't do anything. They're not "real" frames, which means there's no decrease in input latency, and you can't see them either. In fact, you might be lowering your actual frame rate in pursuit of a high "total" frame rate.

All of this is, by the way, is without even getting into artifacting issues that seem to differ on a game-by-game basis, where some titles appear to really struggle with frame generation more than others. Alan Wake II is a great example here as well, as it's a game that you navigate with a torch. Around the edge of the illuminated field and even in the center of your illuminated area can frequently look strange and have significant artifacting thanks to multi

frame generation.

New tech is always exciting, and some of the other improvements in DLSS 4 are very, very interesting, but multi frame generation is a bit of a headscratcher in my eyes. When it comes to gaming, there are very niche use cases where it actually makes sense and will give a significant benefit to the player. These fundamental latency-related problems are also likely impossible to solve, which means that there are very improvements that can be made to overcome the issues that I (and countless others) have outlined.