Nvidia Reflex is a rather ambiguous technology that's ultimately marketed as a tool designed to minimize the delay between your mouse and keyboard input followed by the action showing up on screen. It's deemed to be particularly important for fast-paced competitive games where every millisecond counts, namely competitive titles like Counter-Strike 2, Valorant, and Battlefield 6. Yet it's a feature that's present in more games than just fast-paced, competitive FPS titles.
As a result, you'd be forgiven for not really know what it actually does. Plus, what on Earth is "Reflex + Boost"? It can be hard to know when to enable either of these settings or to figure out what they actually do, especially given that the feature is often buried in a video setting labeled "Reflex" without any other context. If you've ever wondered what exactly Reflex is, along with what it does, it can be hard to read through all of the developer documentation to get a clear picture of what's happening when you enable it.
As for what it actually does to reduce latency under the hood, the answer is, well, a lot. It's vital to understand what Reflex is as a whole before we dive into the "Boost" part of the equation.
What is Nvidia Reflex?
It's been around a long time, now
First, a bit of a primer: Nvidia's Reflex is a low-latency mode that has to be built into a game (so, it needs developer support) specifically leveraging Nvidia GPUs from the GTX 900 series and newer. Its goal is to reduce system latency overall, and can ultimately be measured by the time between a user action (like clicking the mouse) and the resulting pixels updating on your display. In practical terms, Reflex is intended to make your controls feel more responsive by ensuring the GPU and CPU work together more efficiently.
Supported games (mostly competitive shooters and similar titles) expose Reflex as a toggle in their settings. For example, games like Valorant, Counter-Strike 2, Fortnite, Call of Duty: Warzone and many others have a setting dubbed "Nvidia Reflex Low Latency" or something similar to that effect. This usually comes with two options: On and On + Boost (more on "Boost" in a bit). If a game doesn't natively support Reflex, Nvidia also provides a driver feature called Ultra Low Latency Mode (NULL) as a fallback, though Reflex integrated in-game tends to be more effective.
But what causes the latency in the first place? In a normal game rendering pipeline, the CPU prepares frames and the GPU renders them, often asynchronously. The CPU might be a few frames ahead of the GPU, meaning it has queued up work for the GPU. If the GPU is the slowest part of the pipeline (such as in a scenario where the GPU is fully utilized), a queue of pending frames builds up. Each frame sits in line waiting to be rendered, which adds delay. That delay means the frame you see on screen might be showing older input or game state. High frame rates can mask some of this, but whenever the GPU can't keep up with the CPU, render queue latency increases. This is, at least somewhat, unrelated to your GPU's overall performance.
Nvidia Reflex addresses this by actively coordinating the CPU and GPU. With Reflex switched on, the game will only submit a frame to the GPU at the last moment when the GPU is ready for it. In other words, the CPU's work is timed "just in time" for the GPU. This prevents the GPU from having a backlog of queued frames, and the CPU effectively waits if necessary so that it doesn't get too far ahead of the GPU. By doing this, Reflex eliminates the excess render queue and thus reduces the latency that queue would cause. The game always renders the newest possible input because the CPU is sampling player input right before sending work to the GPU for that frame.
The end result is pretty impressive, and third-party tests have shown a considerable decrease in input latency with the feature enabled. Reflex works best when the GPU is the limiting factor, as that's when frame queues would normally build up. In a situation where the GPU is the bottleneck, Reflex can dramatically lower end-to-end latency by keeping the pipeline short. In CPU-bound scenarios (where your CPU is the bottleneck and your GPU is underutilized), standard Reflex may not provide as much benefit. After all, if the GPU isn't being fed faster than it can handle, there's no long queue to cut down on. In those cases, latency is already low to begin with, as it's limited by how fast the CPU produces frames.
If your game of choice doesn't support the Reflex SDK, you still have another option, and that's to enable NULL. Open the Nvidia Control Panel, navigate to "Manage 3D Settings", then select "Low Latency Mode", and choose the Ultra option. It will help reduce latency, but won't provide full control of the rendering pipeline to your GPU.
How Nvidia Reflex actually works
It ties your CPU and GPU together
So, you know what Nvidia Reflex is, but how does it actually work? Under the hood, it makes use of both the game engine and the GPU driver working in concert in order to decrease end to end input latency. The Reflex SDK, which game developers integrate, provides an API to manage frame submission timing. Essentially, the GPU driver provides feedback to the game when the GPU is ready for more work. The game can then delay its next frame's processing until just before that moment. It's a constant communication between the GPU and CPU, and Reflex is checking to make sure that the CPU doesn't go too far ahead.
In a normal scenario without Reflex, and the GPU is struggling to keep up with what the CPU is preparing, you might find that the GPU is a couple of frames behind what the CPU has already completed the work for. Those frames sit in a render buffer on the GPU, and it means that input or updates that happened after those queued frames won't be seen until the queue clears. For example, let's examine the following:
- Your CPU has prepared and queued N frames for rendering
- The GPU is operating on N-3 frames (in other words, three frames behind)
If you click your mouse at time T (frame N on the CPU), but the GPU is N-3 frames behind, your click won't actually appear on the screen until T+3, which is when your GPU reaches that Nth frame. Reflex's "just-in-time" scheduling ensures that when you click at time T, the very next frame the GPU renders (if at all possible) is the one that includes that click, rather than being stuck behind previously queued frames.
All of this is why Reflex can even half your input latency in some cases. In a game where your frame rates are incredibly high, there's almost no difference, but intensive games or games that are running at a lower FPS will see dramatic improvements. For example, Valve recommends playing Counter-Strike 2 with Gsync and Vsync switched on (where Vsync would normally incur an input latency penalty as it forces the GPU to hold back frames until the monitor is ready for them), but the combination of Gsync and Reflex + Boost eliminates it entirely, as the next frame is delivered as soon as possible.
There's one downside to Reflex, though it can be argued that it's not really a downside, and more of a side-effect of getting more consistent inputs. By keeping the pipeline as short as possible, Reflex might occasionally result in the CPU not being 100% utilized if it's waiting for the GPU. There could be a minor trade-off in absolute throughput (FPS) in some cases, since Reflex prioritizes freshness of frames over sheer number of frames rendered. In practice, if you have excess frame rates, a slight frame rate reduction is usually a worthwhile trade for lower input lag. If you're already at low frame rate, Reflex doesn't magically raise it (it's not a performance boost feature), but it ensures you aren't getting additional lag on top of a reduce frame rate.
The "Boost" in On + Boost is mostly unrelated
Though it still has benefits
All of that leads us to the final piece of the puzzle: Nvdia's "On + Boost" Reflex setting. The Boost part is an extra feature that goes beyond the standard latency reduction logic that Reflex uses. It primarily affects the behavior of the GPU, specifically its clock speeds and power state, in order to minimize latency further in certain situations.
Modern GPUs have power management that can downclock or idle at lower frequencies when they aren't fully utilized (to save energy or reduce heat). In a scenario where the game is CPU-bound (the GPU isn't working hard and is waiting on the CPU), then the GPU might lower its clocks to save power. Normally this is fine, but if suddenly the CPU finishes a frame and hands a bunch of work to the GPU, a downclocked GPU may take a few extra milliseconds to ramp up and complete that work. Those extra milliseconds add latency, and can cause sudden frame rate dips.
Reflex + Boost prevents the GPU from going into a power-saving, low-clock mode during gameplay, even if the scene is easy to render in the moment. When Boost is enabled, the GPU will stay at higher clocks or ramp up more aggressively instead of idling down, even if the CPU is the limiting factor at that point in time. Essentially, Boost is overriding the GPU's power management to keep it at full throttle all of the time, so it's always ready to render the next frame as fast as possible. Nvidia states this can "reduce latency slightly when the GPU is significantly underutilized." The trade-off, as you might guess, is higher power consumption and potentially more heat. Your GPU might run louder or use more power with Boost on, because it's not allowed to downclock as much.
It's important to make clear that Boost does not make your GPU faster than its normal capability. It doesn't overclock beyond its standard boost clocks, and all it does is ensure that the GPU is at full performance when needed, rather than briefly going to an idle or lower-power state. In scenarios where the GPU is already fully utilized (like running at a high quality graphics preset and the GPU is already at full pelt), then it's likely operating at a high clock speed anyway. In other words, Boost won't provide much (if any) benefit on its own. After all, there's no headroom to behave differently, and the GPU is already at max performance.
This has benefits in games that aren't as taxing on your GPU, and Counter-Strike 2 is a great example of that. By keeping clock speeds high, a sudden flurry of grenades, molotovs, and smoke grenades won't cause a severe frame rate dip (assuming you also play with Vsync and Gsync switched on), as your GPU is already running at full power and ready to go. In fact, the Reflex + Boost in-game setting is analogous to manually setting your GPU to a "Prefer Maximum Performance" mode in the Nvidia Control Panel.
When should you enable Nvidia Reflex?
All the time, basically
Nvidia Reflex is a tool that you should always enable when you want the lowest possible latency. It manages your CPU and GPU in a way that allows them to work together, so that frames are produced just in time and are based on the latest input, rather than being queued up. The standard Reflex mode already brings significant latency reductions in GPU-bound cases by eliminating render queue bottlenecks. The Boost option dials things up a notch, by also ensuring that your GPU is always at the ready during gameplay to deal with a sudden spike in objects to be rendered and calculations to be done.
Really, you should enable Reflex in every supported game that you want the fastest response times possible in. There's essentially no downside, and the upsides include better reaction times and potentially a competitive edge. If you're on a desktop, you can enable Boost as well... so long as you don't mind the extra power usage. It's most useful when your GPU headroom is high and you want to ensure there are no slowdowns. In full GPU load situations, Boost may not change much, but Reflex itself will achieve a lot.
In competitive games like Valorant or Counter-Strike 2, Reflex helps maintain consistent low latency at high frame rates. And in graphically intense games, it prevents the added lag that usually comes with maxing out your GPU. Across the industry, technologies like AMD's Anti-Lag 2 and Intel's Xe Low Latency are built on the same principle of aligning the CPU and GPU's work in order to reduce input lag. Nvidia's Reflex with Boost currently remains the gold standard largely due to its widespread industry adoption and proven results in many titles, but it's undoubtedly a good thing to see the competition pushing similar features for all gamers.
At the end of the day, if responsiveness matters to you, Nvidia's Reflex (and its counterparts on other hardware) are well worth turning on. Unless the game itself has a poor implementation of the feature, there's no downside, and you're choosing to leave a potentially better experience on the table with it switched off.