As my exploration of Frame Generation explains, I’m quite skeevy of Nvidia’s all-in approach to AI-based performance modifiers. Baseline DLSS is solid, especially now that DLSS4 is here, but the rest I’m iffy on. What about features such as Nvidia Reflex, then, where the boons are even harder to spot?
Nvidia Reflex has felt oblique to me from day one. Whereas the actual super-sampling part of DLSS has always been rather easy to identify through the use of one’s eyes, this has never been the case with Reflex. Heck, even the infamous Frame Generation component of Nvidia’s DLSS package is easy to notice when you crank it up to eleven. As far as Reflex goes, though, the situation is meaningfully different.
As one of our readers suggested a little while ago (thanks, Icy!), a feature explaining the ins, the outs, and the what-ifs of Nvidia Reflex has been a long time coming, and this is precisely what I’ve set out to cook up here. If you, too, are having trouble figuring out what on earth Reflex is even doing, this is the place to be.
What is latency, and how does Nvidia Reflex affect it?
In the simplest terms possible, Nvidia Reflex is a feature suite designed specifically to combat click-to-display system latency. Latency is the exact delay between your input and the appearance of its effects on-screen, and the higher it gets, the less responsive your game will feel. If you’re looking for a handy little cheat sheet to keep track of, that’s easy:
- High frame rate and low latency are good.
- Low frame rate and high latency are bad.
- A low frame rate can lead to high latency, but it’s usually just one of its notable components. You can have a high frame rate with high latency (i.e., Frame Generation usage).
Older gamers may also know latency under its funnier-sounding moniker of “ping,” and this, too, may help you start connecting the dots in real time. Remember playing Modern Warfare on a shoddy Internet connection back in 2008? Yeah, not the kind of thing you just forget, is it now? Back on track, though.
Once upon a time, frame rate was the go-to performance quantifier for video games. The higher the number, the better your FPS, and the better your FPS, the more responsive your games felt. Easy, right? The thing is, we’ve since complicated things a fair bit by driving home the fact that, no, FPS isn’t the only thing we should be looking at in this context. Instead, we’ve got the frame rate itself, frame rate consistency, and input latency to take into account.
The overall perceived performance of a game (how it both looks and feels to play in action) is a combination of its FPS and input latency, among other things, and Nvidia Reflex attempts to reduce the latter part of the equation as much as possible. Sometimes, it is to the user’s detriment, but more on that later.
How does Nvidia Reflex work?
Nvidia Reflex sets out to “optimize every aspect of the rendering pipeline for latency using a combination of SDKs and driver optimizations,” as per Nvidia’s blog post from a few years back. For maximum effect, Reflex needs to be implemented into games on a per-project basis for it to work, and that’s because it’s egregiously difficult to reduce input latency otherwise.
The Reflex Software Development Kit (SDK) fine-tunes a game engine to achieve just-in-time rendering: as soon as a frame is rendered, it gets pushed straight to the screen to reduce the CPU’s render queue for the GPU. The traditional rendering pipeline works by having the CPU queue up frames for the GPU to render, and Reflex effectively shortens this window as much as it possibly can.
“When developers integrate the Reflex SDK, they are able to effectively delay the sampling of input and game simulation by dynamically adjusting the submission timing of rendering work to the GPU so that they are processed just-in-time,” Nvidia explains. And oh, hey, it usually works out just fine.
On a practical level, Nvidia Reflex makes it so that your GPU always stays under 100% usage, which leads to a slight potential reduction in frame rate while massively reducing input delay. If your GPU gets choked up, you’ll get extreme stuttering and other performance problems, which then leads to a greater latency than you get by keeping your GPU under its maximum load capacity.
In traditionally rendered games with the Reflex SDK fully enabled, you will get measurable input latency improvements, and that genuinely is what it’s all about. As a result, you get a game that’s substantially more responsive and feels better to play, which is a huge deal. Nvidia Reflex isn’t a silver bullet of any kind, though.
Problems with Nvidia Reflex and its use cases
Even though it can seem like magic, Nvidia DLSS as a feature-suite is anything but. I already mentioned the importance of keeping system latency in mind when optimizing video games, but there’s one more thing that we need to drive home here: latency cannot be entirely nullified. Processing takes time as a fact, and the more stuff you pile into the rendering pipeline, the more latency you are adding to the render equation. With that in mind, it rubs me the wrong way that Nvidia is kinda-sorta pitching Frame Generation as the ultimate solution to the games’ performance problems.
Frame generation of any sort improves the perceived visual smoothness of a game: its frames-per-second. It does absolutely nothing as far as latency is concerned. No, actually, scratch that: frame generation has a rendering cost attached to it, which means it adds a few milliseconds to your rendering pipeline as a rule. There’s no getting around that, though Nvidia Reflex certainly attempts to do so.
Make no mistake: using Nvidia Reflex alongside Nvidia Frame-Gen is a huge help for input latency, but you’re still getting 60 FPS’ worth of ping at 200 FPS’ worth of perceived smoothness. If that sounds like a bit of a mismatch, you’re on the right track.
Aside from that, Nvidia Reflex can generally break down at times. In some games, it very well will lead to reduced performance, unstable frame-pacing (a huge no-no in my book), and even visible stuttering. There’s a reason why you don’t want to just enable Reflex at a driver level and go with it and why Nvidia recommends that the developers use Reflex SDK for proper implementation.
How to use Nvidia Reflex properly?
Even though Nvidia Reflex is a complicated beast in its own right, using it is quite the opposite: just enable the Nvidia Reflex Low-Latency Mode option in your game’s graphics settings, and you’re golden. Once again, note that this incurs a performance penalty by default, though this will be a net positive for your input latency in the vast majority of cases.
Nvidia Reflex On vs On + Boost
The more complicated question concerns the fact that there are usually two different versions of Nvidia Reflex to enable in games: Reflex On and On + Boost. There is a meaningful difference between the two, too.
The baseline Reflex essentially maintains a dynamic frame-rate cap that keeps your GPU under 100% load to improve input latency. The On + Boost option, on the other hand, attempts to crank your GPU’s core frequency to improve your 1% lows.
If you’re unfamiliar with the verbiage, 1% lows are an extremely important frame-rate statistic that has to do with that third element of video game performance I had previously only glossed over: consistency. 1% lows are the worst, slowest frames your GPU can render, and in a perfect scenario, they’d be as close to your actual average FPS as possible. The bigger the delta between your average FPS and your 1% lows, the worse your frame-pacing will be.
With this information in tow, Nvidia Reflex On + Boost tries its darndest to boost your 1% lows by forcing your GPU to run at a high frequency. When your GPU runs fast, however, it heats up more quickly as well. What happens when your GPU overheats? It drops frequency to cool off. You can see now why On + Boost can lead to some performance problems that don’t seem obvious at a glance.
It’s a great idea, to be sure, but whereas I’m happy to recommend you keep Reflex enabled wherever possible, I would advise testing for the On + Boost option just to be safe.
Nvidia Reflex is a good feature, but keep its caveats in mind
There’s an awful lot to keep in mind about Nvidia Reflex because it’s a complex technology that attempts to solve a complex problem. The bottom line is thankfully rather simple, though: if a game has the option for you to enable it, you should do so.
The flip side of this particular coin is that Reflex will most certainly not help Multi-Frame Generation feel like you’re getting a native 400 FPS’ worth of performance. It feels to me like Nvidia is trying to prime people to think that Reflex is something it’s really not. Reflex 2, the upgraded version of the stuff I’ve discussed above, looks great as well, but it won’t be the magical solution to the hard problem of latency. Latency is here to stay, and Frame Generation only adds to it.
Published: Jan 26, 2025 09:25 am