Nvidia’s DLSS upscaling technology is the most exciting thing happening in gaming that sounds like the least exciting thing. I’ve been crowing about what an amazing tool DLSS is for years, going all the way back to when we thought the Switch 2 was just the Switch Pro.
I mean, I guess it kind of is… but I digress.
DLSS 4 released alongside the newest generation of 50-series GPUs, and I’ve been putting some of the new features through their paces to find out if they’re everything Nvidia says they are. In particular, I was interested to see how multi-frame generation, which now supports 4x on 50-series cards, can be used to improve performance and visual fidelity beyond my system’s normal limits. After much testing, it’s safe to say the marketing isn’t deceptive: DLSS 4’s multi-frame generation is so good at creating frames out of thin air, it practically looks like magic.
You Get Some Frames, And You Get Some Frames
A lot of graphics and gaming tech is hard to parse because it sounds like a bunch of technobabble, but multi-frame generation is exactly what it sounds like. While DLSS, or deep-learning super sampling, boosts performance by running your game at a low resolution using AI to infer a higher quality image, multi-frame generation uses AI to, well, generate more frames. Nvidia gets into the nitty-gritty of how it’s able to do that here, but what I’m really interested in is whether it or not it actually works, and what drawbacks, if any, it comes with.
To test multi-frame generation, Nvidia provided me with an RTX 5080, Doom: The Dark Ages, and Dune: Awakening, to see how the feature works in both single player and multiplayer settings. There are already 100 games that support multi-frame generation, with new releases like Mecha Break, Phantom Break Zero, and Tides of Annihilation on the way, but between Doom and Dune, I got a pretty good sense of what DLSS 4 has to offer modern games. I also tested God of War Ragnarok and Spider-Man 2 to see what the tech does for slightly older games too.
RTX 40-series cards don’t support multi-frame generation, but still offer 2x frame generation.
Doom: The Dark Ages is one of the most frenetic shooters out there, so multi-frame generation would have to be pretty precise to keep up with the Slayer’s rampage. As a basic test I started with DLSS in balanced mode and tuned my settings until I got a stable 60fps. I use a 5120×2160 super ultrawide display, so even with a 5080 slotted in, locking in at 60 is a lot easier than it sounds.
On 50-series cards multi-frame can scale from 2x to 4x, so I turned on 2x and voila, 120fps. I skipped 3x and went right to 4x, and what do you know, 240fps. One click of a button, and my framerate quadrupled. Could this be true? Is it really that simple? Well, yes and no, but mostly yes.
Could It Really Be That Easy?
Going from 60fps to 240 with the click of a button is pretty incredible. As someone who has been PC gaming since the ‘90s and has spent hours tweaking shadow settings to squeeze two extra frames out of my rig, I couldn’t believe how easy this was. No really, I couldn’t believe it.
I played hours of Doom: The Dark Ages, and my conclusion is that multi-frame generation is incredible, but it’s not perfect. You will find some occasional artifacting if you’re looking for it, especially in things that are difficult to render like detailed foliage or when objects move out from behind UI. Even around the target reticle, which is where your focus is when locked when you’re playing, you can see some of the calculations happening in front of your eyes, depending on what you’re aiming at. The flaws are the most notable on 4x, while on 2x, I barely noticed any flaws at all.
The other thing to consider is that using multi-frame generation increases latency. This won’t be a huge deal in some games, but for something super twitchy like Doom, it can feel a little bit floaty. The latency didn’t bother me nearly as much in Spider-Man or God of War, but I definitely noticed it in Doom. While getting lots of free frames may sound great for competitive shooters, because of the latency, I wouldn’t recommend this feature for Marvel Rivals or Apex Legends players.
But for something like Dune: Awakening, it’s incredible. The smoothness you get from having a higher frame rate has a big impact. Driving and flying across the desert is that much more immersive, and the scale of the world, along with its muted color palette, makes any potential flaw a lot harder to detect. Of all the games I tested, Dune was the most impressive.
Who Is Multi-Frame Generation For?
My first instinct when I learned about multi-frame generation was that, like DLSS itself, it would help future proof our gaming PCs. It doesn’t matter if your hardware is too low to hit a stable 30 when you can hit one button and turn 30 into 120. It turns out that, at least for now, low end machines won’t get as much out of multi-frame generation as machines that can already hit a decent frame rate. Going from 30 to 120 introduces significantly more artifacting than going from 60 to 240, simply because your starting point is so much lower.
That doesn’t mean it’s useless for low-end machines, and different people will have different sensitivities to defects. There’s a good chance you will be able to squeeze more life out of lower end 4060 and 5060 GPUs thanks to this feature, because the benefits will be worth more than the sacrifices.
The best use for multi-frame generation seems to be maxing out the refresh rate of your display. My monstrous super ultrawide has a 240Hz refresh rate that I’ve never been able to take advantage of, but with multi-frame generation, I can finally get there. It’s not uncommon for gamers to have 120, 240, or even 360Hz displays these days, and for modern triple-A games, this is the only way you’re going to be able to hit those kinds of numbers.
Is multi-frame generation a magical make more frames button? Well yeah, it kind of is. When you consider the drawbacks, it’s pretty impressive how much more you can get out of your rig. After three years with the Samsung Odyssey G9, I’m finally getting to see its full potential thanks to this feature, and this is only the start.
- Date Founded
-
April 1, 1993
- CEO
-
Jensen Huang
- Subsidiaries
-
Mellanox Technologies, Cumulus Networks, NVIDIA Advanced Rendering Center
- Headquarters
-
Santa Clara, California, United States