NVIDIA Reflex 2 and Frame Warp Explained: Real Input Lag Cut

NVIDIA Reflex 2 is a two-part latency tech — the original Reflex Low Latency mode plus a new step called Frame Warp — that NVIDIA measures at up to a 75% cut in PC latency versus no Reflex, by re-projecting the finished frame to match your latest mouse input right before it hits the screen. It reduces input lag (unlike frame generation, which adds it), it debuts first on RTX 50 Series cards, and as of mid-2026 it is still not broadly released. This guide explains what Reflex 2 and Frame Warp actually do, the real numbers, how they differ from frame generation, and what to do for lower latency right now.

What Reflex 2 is

Reflex 2 stacks two techniques that both attack system (PC) latency — the time from your mouse move to the pixels changing on screen:

1. Reflex Low Latency (the original Reflex, shipping since 2020). It keeps the render queue short so the GPU starts each frame with the freshest input, and lets the CPU sample your mouse later, closer to when the frame is submitted. This alone typically cuts latency substantially in GPU-bound games.
2. Frame Warp (new in Reflex 2). After a frame is fully rendered, Frame Warp shifts that image to match your most recent mouse and camera position, sampled at the last possible moment before display.

The pitch: Reflex 1 makes each frame start with newer input; Frame Warp updates the frame with even newer input after it is already drawn.

How Frame Warp works

Frame Warp takes the last rendered frame and re-projects it using the newest camera position derived from your latest mouse input. Because the pixels are pushed to match where you are now aiming — not where you were aiming when rendering began — small gaps open at the edges of the moved image (areas the original frame never drew). NVIDIA fills those gaps with a latency-optimized predictive in-painting algorithm that uses camera, color, and depth data from previous frames to reconstruct the missing slivers.

The effect is that the on-screen image reflects input from a moment later than the frame was rendered, cutting the last stretch of the mouse-to-photon pipeline that Reflex 1 can’t touch.

The latency numbers

NVIDIA’s headline figures come from two demos. Treat them as best-case, vendor-measured results — your mileage varies with GPU, resolution, and CPU/GPU load — but the direction is real:

Scenario (NVIDIA demo)	No Reflex	Reflex (Low Latency)	Reflex 2 (Frame Warp)
THE FINALS, RTX 5070, 4K, max settings	56 ms	27 ms	14 ms (−75%)
VALORANT, RTX 5090, 800+ FPS	—	—	under 3 ms avg PC latency

The key read: most of the first drop (56 → 27 ms) is ordinary Reflex 1, which you can already use today. Frame Warp is responsible for roughly the second halving (27 → 14 ms). NVIDIA says the technique helps in both CPU-bound and GPU-bound situations.

Reflex 2 vs frame generation — opposite effects on lag

This is the point most people get backwards, so it earns its own table. Frame generation and Frame Warp both involve manipulating frames, but they move latency in opposite directions:

	Frame Generation (DLSS FG / FSR 3)	Reflex 2 Frame Warp
What it does	Inserts a new interpolated frame between two real ones	Re-projects a real frame to match newer input
Goal	Smoother motion / higher frame counter	Lower input latency / faster aim response
Effect on input lag	Adds latency (holds a frame back)	Reduces latency
Carries your input?	No — generated frames add no responsiveness	Yes — updates the frame with newer input
Best for	Single-player, high base FPS, filling a high-refresh panel	Competitive shooters where response matters

If you want the full mechanism behind why generated frames add lag, see Frame Generation and Input Lag Explained. The short version: more frames on the counter is not the same as faster input, and only Frame Warp among these actually lowers the number that matters for aim.

Availability in 2026 — don’t wait for it

Frame Warp was unveiled at CES 2025 and announced to arrive first in THE FINALS and VALORANT, debuting on GeForce RTX 50 Series GPUs, with support for other RTX cards planned afterward. As of mid-2026, however, it still has no broad public release — RTX owners are largely still waiting, and neither launch title has shipped a publicly available Reflex 2 implementation yet.

So Reflex 2 is a real technology with impressive demos, but it is not something to plan your setup around today. The honest move is to cut the latency you can cut right now:

• Turn on the original Reflex (or NVIDIA App’s low-latency option) in every game that supports it — that’s the 56 → 27 ms half of the gain, available today. See Should You Turn On NVIDIA Reflex?.
• Cap your FPS correctly so you stay inside your monitor’s refresh window and out of the GPU-bound queue that inflates latency: How to Cap Your FPS Correctly.
• Clean up the rest of the input pipeline — polling rate, background load, and a lower timer resolution all shave milliseconds before any frame is drawn: How to Minimize Input Delay for Competitive Gaming.

The free Tier1Suite bundles our timer-resolution, tuning, and monitor tools to lower system latency now, no new GPU or unreleased feature required.

Bottom line

Reflex 2 is worth understanding because it does the one thing frame generation can’t: it makes a rendered frame reflect newer input, cutting latency rather than adding it, for up to a 75% total reduction in NVIDIA’s demos. But it’s still gated to RTX 50 and largely unreleased in mid-2026, so it changes nothing about your setup today. Turn on Reflex 1, cap your frames, and tighten your input pipeline — those are the milliseconds you can claim right now, and they compound with Reflex 2 whenever it finally ships.

Related guides

• Should You Turn On NVIDIA Reflex?
• NVIDIA Reflex vs AMD Anti-Lag vs Intel XeLL
• Frame Generation and Input Lag Explained
• G-Sync, V-Sync and Reflex Low Latency Setup
• How to Minimize Input Delay for Competitive Gaming
• How to Cap Your FPS Correctly