NVIDIA Reflex is a set of technologies, built into games through an SDK and supported by GeForce drivers, that reduces system latency by synchronizing CPU work with the GPU so that frames do not wait in a render queue. In gaming performance, it addresses the gap between pressing a button and seeing the result on screen, which matters most in competitive shooters and esports titles.
Reflex is not a frame rate booster. It changes when the game samples input and starts each frame, so the same frame rate can feel noticeably more immediate. The family also includes latency measurement tools and, in its second generation, a technique that adjusts the finished frame to the newest mouse movement just before display.
How NVIDIA Reflex removes the render queue
Reflex works by pacing the game engine so the CPU submits each frame just in time for the GPU to render it. In a typical GPU-bound game, the CPU runs ahead, prepares several frames and parks them in a queue. Each queued frame was built from input sampled earlier, so the image on screen reflects old mouse positions.
With Reflex enabled, the engine calls the Reflex SDK at defined points in its frame loop. The driver measures how long the GPU needs per frame and tells the engine to delay the start of CPU work until the GPU is nearly ready. The queue stays close to zero frames, and input is read as late as possible. The table below lists the stages of one frame and how Reflex affects each.
| Stage | What happens | Effect of Reflex |
|---|---|---|
| Input sampling | Game reads mouse and keyboard state | Moved later, closer to rendering |
| Simulation | CPU updates game logic and physics | Unchanged in duration |
| Render submission | CPU builds draw calls for the GPU | Timed to match GPU readiness |
| Render queue | Finished CPU work waits for the GPU | Largely eliminated |
| GPU rendering | GPU draws the frame | Unchanged, or clocks held high with Boost |
| Composition and scanout | Frame is presented and sent to the display | Unchanged |
Reflex modes and components
Reflex is a family of related features rather than a single switch, and each part solves a different problem. Games expose the low latency mode in their settings menu, while the analysis tools live in monitors, mice and overlay software.
| Component | Purpose | Where it runs |
|---|---|---|
| Reflex Low Latency: On | Keeps the render queue empty | Game engine and driver |
| On + Boost | Keeps GPU clocks high when CPU-bound | Game engine and driver |
| Reflex Latency Analyzer | Measures click-to-photon latency | Supported monitor and mouse |
| PC Latency statistics | Reports render latency in overlays | Game markers and overlay software |
| Reflex 2 Frame Warp | Shifts the final frame to the newest camera input | Game engine and GPU |
The Boost option exists because in CPU-bound scenes the GPU idles between frames and may lower its clock speed, which lengthens the next frame. Boost trades higher power draw for steadier GPU timing. Frame Warp in Reflex 2 goes further: after rendering, it reprojects the image using the latest mouse data and fills the uncovered edges with predicted pixels, so camera movement on screen reflects input sampled only a few milliseconds earlier.
How Reflex compares with related latency techniques
Reflex is most often compared with driver-level low latency settings, frame rate caps and competing vendor technologies. The key difference is that Reflex runs inside the game engine, so it controls input timing directly rather than guessing from outside.
| Technique | Mechanism | Main limitation |
|---|---|---|
| NVIDIA Reflex | Engine-side just-in-time frame submission | Requires integration by the developer |
| Ultra Low Latency Mode | Driver limits queued frames | Cannot move input sampling inside the engine |
| Frame rate cap | Limits frames so the GPU is never saturated | Wastes headroom and needs manual tuning |
| AMD Anti-Lag 2 | Engine-integrated pacing on Radeon GPUs | Supported in fewer titles |
| Intel XeLL | Engine-integrated pacing on Arc GPUs | Newer and less widely adopted |
Reflex also interacts with variable refresh rate. When G-SYNC and V-Sync are both enabled, Reflex applies an automatic frame cap slightly below the refresh rate, which keeps the game inside the VRR window and avoids the latency penalty of V-Sync back pressure. Frame generation in DLSS 3 and later requires Reflex, because generated frames add delay that Reflex partly offsets.
When Reflex helps and how much
The benefit of Reflex depends mainly on whether the GPU is the bottleneck. In heavily GPU-bound games the queue grows to one or more full frames, and Reflex can reduce PC latency by roughly a third to a half. When the CPU is the limit, the queue is already short and the gain is small.
| Scenario | Typical gain | Recommended setting |
|---|---|---|
| GPU-bound, high settings | Large reduction | On |
| CPU-bound, low settings | Small reduction | On + Boost |
| Frame generation enabled | Offsets added delay | On, enforced by the game |
| VRR with V-Sync | Moderate, avoids V-Sync queue | On with G-SYNC |
| Very high fixed cap | Minimal | On, little downside |
Exact numbers vary with the engine, resolution and hardware. A practical rule is that Reflex costs almost nothing in frame rate, often less than a few percent, so enabling it is the sensible default in any supported game. GeForce GTX 900 series and newer cards support the low latency mode.
Limitations and common misconceptions
Reflex reduces only one part of end-to-end system latency. Peripheral delay from the mouse and display processing in the monitor remain, and a slow panel can add more delay than the render queue ever did. Several misunderstandings appear regularly among players.
- Reflex does not raise frame rate; it may lower average frames per second slightly while improving responsiveness.
- Driver settings cannot replace Reflex; Ultra Low Latency Mode is a fallback for games without integration.
- Network latency is untouched; netcode delay and server tick rate are separate from local rendering delay.
- Frame Warp changes only camera movement; animation and hit registration still follow the rendered frame.
- Boost increases power and heat, which matters on laptops and small form factor systems.
Explained without jargon
Imagine a kitchen where the chef prepares several plates in advance and leaves them waiting. By the time a plate reaches the table, the order may have changed. NVIDIA Reflex tells the chef to start each plate only when the waiter is ready, so every dish reflects the latest order.
In a game, this means your mouse movement and clicks appear on screen sooner. The picture does not get smoother or prettier, but the game feels quicker to respond, which helps most when split-second reactions decide the outcome.
A typical use case
A competitive player runs a tactical shooter at 1440p on a high-end GeForce card with a 240 Hz G-SYNC monitor. With maximum graphics settings, the GPU is fully loaded and the game reaches about 180 frames per second, yet aiming feels slightly heavy. An overlay showing PC latency reports around 30 ms.
After switching Reflex to On, the frame rate drops by only a few frames, but the overlay now shows PC latency of about 17 ms. Because G-SYNC and V-Sync are active, Reflex also caps the game just below 240 frames per second. The player notices that quick flicks land more precisely, and the improvement comes without buying new hardware.