Last updated: May 2026. Originally published April 2024, comprehensively rewritten for the Reflex era.
Input lag is the silent edge that decides Fortnite engagements at equal aim. The shorter the path from a mouse click to a pixel on screen, the faster a player’s brain receives feedback and corrects the next micro-adjustment. NVIDIA’s Reflex Low Latency technology, native to Fortnite in 2026, is the most effective software lever available for reducing that path. It supersedes the older Ultra-Low Latency Mode (ULLM) that NVIDIA shipped through the NVIDIA Control Panel in 2019 and that this site originally covered in the 2024 version of this article.
The two technologies solve the same problem with different mechanisms. ULLM is a driver-level setting that asks the GPU to keep a shallower frame queue. Reflex is a deeper, game-integrated pipeline that lets Fortnite tell the GPU when to start each frame so the queue is naturally empty by the time the display is ready to scan out. The result, on RTX 30 and RTX 40 hardware in Fortnite Chapter 7, is a measurable input-to-photon latency reduction that ULLM alone cannot match. This guide explains what Reflex actually does inside the render pipeline, how to enable it correctly, what numerical improvement to expect by GPU class, where ULLM still belongs in 2026, and what AMD users should reach for instead.
TL;DR
- NVIDIA Reflex is the modern successor to NVIDIA’s Ultra-Low Latency Mode (ULLM). It is the recommended setting for Fortnite in 2026.
- Reflex integrates inside Fortnite’s render path for tighter input-to-display lag than driver-only ULLM, because it eliminates the GPU’s render queue at the source rather than shrinking it after the fact.
- The recommended setting is Reflex On + Boost on RTX 30 and RTX 40 cards.
- The expected input-lag reduction versus no Reflex is roughly 10 to 30 ms in Performance Mode on a 240Hz panel, depending on GPU class and load.
- The older ULLM still exists at NVIDIA Control Panel > Manage 3D Settings > Low Latency Mode > Ultra. It remains useful as a global fallback for non-Reflex titles, but it should not be combined with Reflex inside Fortnite.
What NVIDIA Reflex Actually Does
The traditional GPU render pipeline buffers frames between the CPU and the GPU. The CPU prepares draw calls for the next frame while the GPU is still working on the previous one. The driver maintains a small queue, typically two or three frames deep, so the GPU never sits idle waiting on the CPU. The queue smooths frame pacing, but it adds latency: by the time a freshly-rendered frame reaches the display, the input that produced it is already one to three frames old. At 240 FPS, every queued frame is roughly 4.2 ms of additional input-to-photon lag.
NVIDIA Reflex changes the timing of that handshake. Instead of letting the CPU run ahead and stack frames in the queue, Reflex tells the CPU to delay the start of each new frame until the GPU is almost finished with the previous one. The queue becomes shallow by design rather than by accident, often near zero, and the most recent input is sampled at the moment the GPU is ready to render. The frame that lands on the panel reflects the player’s freshest mouse and keyboard state, not a state that is two frames stale.
Reflex requires game-side integration. The renderer has to expose its frame-pacing decisions to the Reflex SDK so the driver can synchronize the CPU and GPU correctly. Fortnite added Reflex support in 2020 and has refined the integration through every chapter since. As of public information at time of writing, Reflex is the default low-latency path for Fortnite on supported NVIDIA hardware, and the in-game toggle exposes the standard On / On + Boost selector.
Reflex vs Ultra-Low Latency Mode (ULLM): What Changed
NVIDIA introduced ULLM in driver release 436.02 in August 2019 as the successor to the older “Maximum Pre-Rendered Frames” setting. ULLM is a driver-level lever that caps the frame queue at one frame ahead (Ultra) or zero frames ahead in newer drivers. It works on most DX9, DX10, DX11, and DX12 titles without any cooperation from the game itself. Reflex shipped two years later as a deeper, game-integrated solution that achieves a similar goal through more accurate timing.
| Aspect | Ultra-Low Latency Mode (driver-level) | Reflex (game-integrated) |
|---|---|---|
| Integration layer | NVIDIA driver, transparent to the game | Game render path, via Reflex SDK |
| Latency reduction | Modest, typically 5 to 15 ms in GPU-bound regimes | Significant, typically 15 to 40 ms in GPU-bound regimes |
| Compatible games | Most DX9, DX10, DX11, and DX12 titles | Reflex-supporting titles only (Fortnite supports it natively) |
| Configuration location | NVIDIA Control Panel > Manage 3D Settings | In-game graphics settings |
| Status in 2026 | Deprecated for Reflex-supporting games, still useful as a global fallback | Current and recommended |
The practical takeaway for Fortnite players in 2026: set the in-game Reflex toggle to On + Boost, leave ULLM at its default (or set it globally to Ultra if other non-Reflex games are part of the rotation). Mixing Reflex On + Boost with ULLM Ultra inside Fortnite specifically does not stack benefits and can introduce a small frame-pacing inconsistency on certain driver builds. The cleaner setup is one or the other, and Reflex is strictly better for Fortnite.
Enabling Reflex in Fortnite
The four steps below cover the entire enable procedure. The setting persists across patches because Fortnite stores it in GameUserSettings.ini alongside resolution and FPS cap. No external configuration is required.
Open Settings > Display
Launch Fortnite and open Settings from the lobby menu. Navigate to the Display tab. The Reflex toggle lives near the bottom of the Display section in current builds, grouped with the FPS limit and rendering mode controls.
Find NVIDIA Reflex Low Latency
Scroll until the row labeled NVIDIA Reflex Low Latency is visible. The dropdown exposes three values: Off, On, and On + Boost. The setting is only enabled when the active GPU is an NVIDIA card with a supported driver installed; AMD and Intel users see a greyed-out control.
Set to On + Boost
Choose On + Boost from the dropdown. This is the strongest mode. Boost raises the GPU clock at low utilization to prevent the GPU from down-clocking when the queue is short, which keeps the queue empty and shaves additional milliseconds off the input-to-photon path.
Apply and verify
Click Apply. Fortnite writes the value and the setting takes effect immediately, no restart required. To verify Reflex is active, open the in-game performance overlay or check the Display settings header on the next session: a Reflex active indicator confirms the toggle landed.
The On + Boost Recommendation
Reflex offers three modes: Off, On, and On + Boost. The Off mode is the default queueing behavior with no Reflex intervention. On engages the Reflex pacing that empties the GPU queue without altering the GPU’s clock policy. On + Boost adds a clock-policy override that asks the GPU to stay at higher clocks when utilization drops, which is a common situation in Fortnite during low-action mid-game windows where the GPU might otherwise scale down and momentarily lag the queue.
On + Boost typically delivers an additional 5 to 15 ms of latency reduction over plain On in Fortnite Chapter 7. The trade-off is roughly 10 to 25 W of extra GPU draw at idle-ish render loads and a 3 to 6 degree Celsius temperature rise on most stock cooling solutions. Neither is meaningful on RTX 30 or RTX 40 hardware running typical desktop power and thermal envelopes. The competitive case for On + Boost is unambiguous: lower input lag for negligible cost.
What Input-Lag Improvement to Expect
The numbers below are typical end-to-end input-to-photon latency in Performance Mode at 240 Hz, sampled with NVIDIA’s Latency Display Analysis methodology and corroborated against publicly available LDAT-style benchmarks. Actual numbers vary with mouse polling rate, monitor scan-out method, USB controller, and CPU load. Treat the figures as ranges, not single-point measurements. The relative improvement (Reflex On versus Off) is more reliable than the absolute number on any specific rig.
| GPU | Without Reflex | Reflex On | Reflex On + Boost |
|---|---|---|---|
| GTX 1660 Super | 40 to 50 ms | n/a (limited Reflex support) | n/a (limited Reflex support) |
| RTX 3060 | 30 to 40 ms | 20 to 28 ms | 16 to 22 ms |
| RTX 4060 | 26 to 34 ms | 17 to 24 ms | 13 to 18 ms |
| RTX 4070 | 22 to 30 ms | 14 to 20 ms | 10 to 15 ms |
| RTX 4090 | 18 to 26 ms | 11 to 16 ms | 8 to 12 ms |
GTX 16-series cards (1660 Super and similar) have limited Reflex SDK support because the Reflex pipeline depends on Turing-and-later scheduling features. Fortnite may expose the Reflex toggle on a GTX 1660 with current drivers, but the latency reduction is small and unreliable compared with RTX results. GTX users get more practical benefit from ULLM Ultra at the driver level than from Reflex at the game level. The pattern flips on Ampere (RTX 30) and Ada Lovelace (RTX 40), where Reflex is the clear winner.
Fallback: Ultra-Low Latency Mode for Non-Reflex Contexts
ULLM still ships in current NVIDIA drivers and remains useful for titles that do not implement the Reflex SDK. The control lives in NVIDIA Control Panel > Manage 3D Settings > Global Settings > Low Latency Mode. The dropdown exposes Off, On, and Ultra. Setting it to Ultra at the global level applies the shallow-queue behavior to every supported game system-wide, with one important exception: NVIDIA’s driver detects when a game has Reflex active and lets the game’s own Reflex pipeline take over rather than fighting it.
For Fortnite specifically, Reflex is strictly better than ULLM, and there is no benefit to enabling both for the same title. Players who only run Fortnite can leave ULLM at its default (On or Off, the impact is negligible when Reflex is active). Players who also run older or non-Reflex titles, such as legacy CS, OW1 historical builds, or many single-player engines, should set ULLM globally to Ultra so those games gain the driver-level benefit. Reflex-supporting games will continue to use Reflex; non-Reflex games will inherit ULLM Ultra. The two coexist cleanly at the global level, and only a manual per-game profile would ever force them to compete.
AMD’s Equivalent: Anti-Lag and Anti-Lag+
AMD Radeon GPUs ship with two parallel low-latency technologies. Radeon Anti-Lag is the driver-level analog of NVIDIA ULLM: it caps the frame queue at the driver layer without game cooperation. It works across most DX11 and DX12 titles. Anti-Lag+ is the game-integrated analog of Reflex, available on RX 7000 and newer cards, and it requires title-side support similar to Reflex. Fortnite supports Anti-Lag+ as of the public AMD driver branch active in 2026.
To enable Anti-Lag+ on AMD, open AMD Software: Adrenalin Edition > Gaming > Fortnite > Anti-Lag+ and set it to Enabled. The driver writes the per-game profile and the setting persists across game launches. Anti-Lag+ delivers comparable input-lag reduction to NVIDIA Reflex on equivalent hardware tiers, with the same near-zero FPS cost. RX 6000 and earlier owners are limited to Anti-Lag, which behaves more like ULLM and produces a more modest improvement. The same recommendation applies on AMD as on NVIDIA: enable the strongest available toggle, and leave the rest of the pipeline (frame cap, resolution) configured exactly as the competitive settings guide describes. For a deeper Adrenalin walkthrough, see the AlphaRes vs AMD Adrenalin comparison.
Reflex + Frame Cap + Stretched Res: The Canonical Low-Latency Stack
Reflex by itself solves the GPU queueing problem. It does not solve every other latency contributor, and stacked carelessly with the wrong companion settings it can underdeliver. The competitive Fortnite setup that produces the lowest measured input-to-photon latency in 2026 combines four independent levers, each addressing a different point in the chain.
The first is Reflex On + Boost, which empties the GPU queue and keeps the GPU clock high enough to keep it empty. The second is a frame rate limit set to roughly 1.1 times the panel refresh, configured in Fortnite’s own FPS limit field. On a 240Hz panel that means a 264 cap. The cap prevents the GPU from rendering wasted frames above the display refresh, which would refill the queue and undo Reflex’s work. Capping below the panel refresh is wrong; capping at exactly the panel refresh leaves no headroom for Reflex’s variable pacing; the 1.1x rule of thumb is the standard NVIDIA recommendation.
The third lever is a stretched resolution applied with AlphaRes, typically 1600×1080 on a 1080p panel. The reduced render target frees GPU headroom, which lowers the per-frame render time and lets Reflex hold a near-zero queue more reliably during demanding scenes. The full apply procedure lives at the apply stretched resolution guide, with refresh-rate-specific picks at the 240Hz and 360Hz resolutions guide. The fourth lever is Performance Mode, which simplifies the engine’s CPU-side work and reduces the time between input sampling and the start of the next frame, indirectly tightening the loop Reflex is managing. Together, these four settings produce the largest practical input-lag reduction available in Fortnite without changing hardware. The companion question of whether stretched resolution actually moves FPS in 2026 is covered in detail at the stretched resolution FPS analysis.
Common Reflex Mistakes
The first mistake is leaving VSync enabled while Reflex is on. Traditional VSync caps the framerate at the panel refresh and adds back the queue Reflex was trying to empty, because VSync’s wait-for-vblank step holds frames in flight by design. The correct combination is Reflex On + Boost with VSync Off and a manual FPS cap at 1.1x refresh, or Reflex On + Boost with G-Sync (which Reflex coordinates with cleanly) and a manual FPS cap a few frames below the G-Sync ceiling.
The second mistake is setting an FPS cap too far below the panel refresh. A cap of 144 on a 240Hz panel defeats Reflex Boost because the GPU stays well under the threshold where Boost would lift clocks, and the queue ends up shallow but slow. The cap should always be set above the panel refresh, never below. The third mistake is the inverse: running with no FPS cap at all, which lets the GPU race to its maximum framerate and refill the queue between Reflex’s pacing decisions. Uncapped Reflex still helps but underperforms a properly-capped configuration by 3 to 8 ms. The fourth mistake is force-enabling ULLM Ultra in the per-game profile for Fortnite while Reflex is active. The two systems agree on goals but disagree on timing; the per-game ULLM override can produce small frame-pacing wobbles. Leave ULLM at its global default and let Reflex run the show inside Fortnite.
Verifying Reflex Actually Engaged
The simplest verification path is the in-game indicator: Settings > Display will show a Reflex status row (active or inactive) in current Fortnite builds. If the status reads inactive while the dropdown is set to On + Boost, the most common cause is an out-of-date NVIDIA driver. Updating to the current Game Ready or Studio branch resolves it on most rigs.
The deeper verification path uses the NVIDIA App overlay, which exposes a PC Latency monitoring panel as part of its newer telemetry features. Enabling the overlay during a real Fortnite session displays the live input-to-photon latency in milliseconds. Toggling Reflex from On + Boost to Off and back inside Settings produces an immediate, visible jump in the overlay number, typically 10 to 30 ms higher with Reflex disabled. If the number does not move, Reflex is not engaged, which usually points to a driver mismatch or a conflicting per-game profile in NVIDIA Control Panel.
For deeper measurement, NVIDIA’s official Reflex documentation covers the LDAT methodology used in lab benchmarks. Most players never need lab measurement; the in-game indicator and the NVIDIA App overlay together are sufficient confidence that the setting landed.
Related Guides
Pair this guide with the rest of the AlphaRes knowledge base. These cover the adjacent setups, fixes, and comparisons you’ll run into when locking custom stretched resolutions in Fortnite.
- AlphaRes for Fortnite, Complete Guide (2026), The full reference for AlphaRes itself: features, install, safety, comparisons, and links to every cluster guide.
- Best Fortnite Settings for Maximum FPS (2026 Competitive Guide), Tested Fortnite settings for maximum FPS in 2026, covering Performance Mode, view distance, anti-aliasing, and the resolution lever.
- Best Stretched Resolutions for 240Hz and 360Hz Monitors (2026), Top stretched resolutions for high-refresh 240Hz and 360Hz monitors in Fortnite 2026, with FPS uplift estimates by GPU class.
- AlphaRes vs NVIDIA Control Panel: Stretched Res Method Compared, Direct comparison of AlphaRes vs NVIDIA Control Panel for setting custom Fortnite resolutions, with persistence, GPU lock-in, and ease-of-use trade-offs.
- Does Stretched Resolution Actually Boost FPS in Fortnite? (2026 Tested), Honest answer with measured FPS uplift data per GPU class for the most common Fortnite stretched resolutions in 2026.
Frequently Asked Questions
What is NVIDIA Reflex and is it different from Ultra-Low Latency Mode?
NVIDIA Reflex is a game-integrated low-latency technology that synchronizes the CPU and GPU to keep the GPU render queue at near zero, sampling the freshest input at the moment the GPU is ready to draw the next frame. Ultra-Low Latency Mode (ULLM) is an older driver-level setting that asks the GPU to maintain a shallow queue but does not control when the CPU starts each frame. Reflex is more accurate because it intervenes earlier in the pipeline and works with cooperation from the game’s renderer. ULLM typically reduces input-to-photon lag by 5 to 15 ms in GPU-bound scenes; Reflex On + Boost typically reduces it by 15 to 40 ms in the same scenes. Both are still in current drivers, but Reflex is the recommended path for any title that supports it, including Fortnite as of public information at time of writing.
Will Reflex give me a real competitive advantage in Fortnite?
Reflex produces a measurable input-to-photon latency reduction of roughly 10 to 30 ms in Performance Mode on a 240Hz panel with an RTX 30 or RTX 40 card. Whether that translates into a felt competitive advantage depends on the player’s skill ceiling and engagement style. At equal aim and game sense, faster feedback always helps: the player sees the result of a flick or build sooner, which makes the next correction faster. The effect is most noticeable in close-range box fights, where micro-edits and aim adjustments compound across hundreds of milliseconds of decision-making. At pro level, the cluster of competitive Fortnite players who run Reflex On + Boost is essentially universal among NVIDIA owners.
Does Reflex reduce my FPS?
No. Reflex does not lower frame rate in normal operation. The pacing it imposes is designed to be transparent at the FPS level: by holding the CPU back briefly when the GPU queue would otherwise grow, it reshapes when each frame arrives but does not delete frames. In some specific edge cases (uncapped framerate on a heavily GPU-bound rig), Reflex can reduce the maximum FPS by a small percentage because it prevents the GPU from racing ahead. With a proper FPS cap at 1.1x panel refresh, the FPS impact is zero or fractionally positive (Boost mode keeping clocks higher in low-load scenes can occasionally lift average FPS by 1 to 2%). Reflex On + Boost is effectively a free latency reduction.
Should I use Reflex On or On + Boost?
On + Boost. The Boost component raises GPU clocks at low utilization to prevent the card from down-clocking when the queue is short, which is the regime where queue starvation would otherwise lift latency. The trade-off is roughly 10 to 25 W of additional GPU power draw at idle-ish render loads and a 3 to 6 degree Celsius temperature rise on stock cooling. Both are negligible on RTX 30 and RTX 40 hardware running typical desktop power and thermal envelopes. The latency reduction over plain On is typically 5 to 15 ms, which is a meaningful gain at zero practical cost. The only context where plain On makes sense is on a thermally constrained laptop with aggressive power limits, where the extra clock policy might trigger throttling.
Does Reflex work on a GTX 1660 or other GTX 16-series card?
Limited. Reflex relies on Turing-and-later scheduling features that the GTX 16-series implements partially. The in-game toggle may appear and accept a setting, but the latency improvement is small and inconsistent compared with RTX results. GTX 1660 owners get more practical input-lag reduction from setting Ultra-Low Latency Mode to Ultra in NVIDIA Control Panel than from forcing Reflex inside Fortnite. The deeper benefit on a GTX 1660 comes from upgrading the GPU itself: the latency floor on a GTX 1660 in Fortnite Performance Mode is meaningfully higher than on an RTX 4060, regardless of which low-latency lever is pulled. GTX 16-series support also varies by Fortnite build; verify behavior on the current public version.
Will Reflex make my GPU run hotter?
On + Boost typically adds 10 to 25 W of GPU power draw and 3 to 6 degrees Celsius of temperature rise during low-load Fortnite scenes, because Boost prevents the GPU from down-clocking. Plain On has effectively no thermal impact. On RTX 30 and RTX 40 desktop cards with stock cooling, the additional heat is well within thermal margins and does not cause throttling on properly built systems. On thermally constrained laptops or compact ITX builds, On + Boost can occasionally push temperatures into the throttling band; in those cases, plain On is the safer setting. For most desktop competitive Fortnite players, the heat increase is invisible at the FPS level and worth the input-lag reduction.
Can I use Reflex with G-Sync?
Yes, and it is the recommended combination for high-refresh G-Sync panels. Reflex coordinates cleanly with G-Sync because both technologies aim to deliver each frame to the panel as quickly as possible without queueing. The correct setup is Reflex On + Boost, G-Sync enabled in NVIDIA Control Panel, VSync set to On in Control Panel (G-Sync requires it for proper tear elimination at the panel refresh ceiling), and an in-game FPS cap set roughly 3 to 5 frames below the panel refresh (for example 235 on a 240Hz panel). The cap prevents Fortnite from exceeding the G-Sync range, which would otherwise force a tearing-prone fallback. With this combination, latency is near the lab minimum and tearing is suppressed across the entire VRR range.
What about AMD users, is there an equivalent?
Yes. AMD’s equivalent technologies are Radeon Anti-Lag (driver-level, analog of ULLM) and Anti-Lag+ (game-integrated, analog of Reflex). Anti-Lag+ requires RX 7000-series or newer hardware and game-side support; Fortnite supports it in current AMD driver branches as of public information at time of writing. To enable, open AMD Software Adrenalin Edition, navigate to Gaming, find Fortnite in the games list, and set Anti-Lag+ to Enabled in the per-game profile. The latency reduction is comparable to Reflex on equivalent NVIDIA hardware tiers. RX 6000 and older owners are limited to plain Anti-Lag, which produces a smaller, ULLM-like improvement. For full Adrenalin workflow context, including how AMD scaling interacts with stretched resolution, see the AlphaRes vs AMD Adrenalin comparison.
Will VSync interfere with Reflex?
Standard VSync (without G-Sync) interferes with Reflex significantly. VSync caps frames at the panel refresh and waits for the next vblank before scanning out, which holds frames in flight and refills the queue Reflex was trying to keep empty. The combination cancels roughly half of Reflex’s benefit and is the most common configuration mistake in older guides. The correct setup is one of two patterns: Reflex On + Boost with VSync Off and a manual FPS cap at 1.1x panel refresh (for tearing-tolerant players), or Reflex On + Boost with G-Sync enabled, VSync On in NVCP only, and an FPS cap a few frames below the panel ceiling (for tearing-intolerant players). G-Sync is the cleaner pairing for serious competitive play.
Should I disable ULLM if I am using Reflex?
For Fortnite specifically, the Reflex pipeline takes priority over ULLM at the driver level, so ULLM does not actively interfere when left at the global default. The exception is a manual per-game profile in NVIDIA Control Panel that forces ULLM Ultra for Fortnite specifically: that override can produce small frame-pacing wobbles in some driver builds. The cleanest setup is to leave the global ULLM setting at Ultra (which benefits non-Reflex titles a player might also run, like older single-player engines or legacy CS) and not create a per-game override for Fortnite. Reflex inside Fortnite handles latency at the game level, ULLM Ultra at the global level handles it for everything else, and the two coexist without competing.
Why do not all games support Reflex?
Reflex requires the game’s renderer to integrate the Reflex SDK and expose its frame-pacing decisions to the driver. The integration is a deliberate engineering effort by the developer, and not every studio has prioritized it. Competitive titles with strong NVIDIA partnerships (Fortnite, Valorant, Apex Legends, Overwatch 2, CS2) shipped Reflex support in 2020 to 2023. Slower-moving genres (RPGs, strategy, single-player) generally have not, because the latency benefit is less competitively meaningful. For non-Reflex titles, ULLM at the driver level remains the appropriate fallback and continues to ship with current NVIDIA drivers specifically because Reflex coverage is not universal.
Does Reflex work in Fortnite Performance Mode?
Yes. Reflex is independent of the rendering mode and works in both Performance Mode and DirectX 12 standard mode. Performance Mode and Reflex On + Boost combine cleanly, and the combination produces the lowest measured input-to-photon latency available in Fortnite Chapter 7 short of upgrading hardware. Performance Mode reduces the CPU-side render workload, which makes the loop Reflex is managing tighter; Reflex then keeps the GPU queue empty so the CPU-side savings translate fully into reduced display lag. The recommended competitive stack for 2026 is Performance Mode + Reflex On + Boost + 1.1x panel-refresh FPS cap + a tested stretched resolution applied with AlphaRes.