LeCoreOS 11 latency investigation — October 15, 2026

Your latest test reports that Fortnite feels delayed. That does not identify a service or registry value as the cause. No measured before/after latency or frame-time capture was supplied. Your screenshots already show driver V-Sync Off, maximum-performance power mode and highest-refresh preference. Low Latency Mode Off is not itself evidence of a fault when the game offers Reflex.

Supported priorities

NVIDIA documents Reflex as coordinating CPU/GPU work and says it supersedes driver Ultra Low Latency. Its guide also says advanced interrupt/scheduling tweaks can make latency worse. This supports testing the game/render path before adding another registry bundle. NVIDIA latency guide.

Epic's current guide recommends DX12 Performance, Low textures/meshes and On + Boost. Its Near view-distance recommendation is changed to Far here because you requested it. Current supported fallback launch flags are -d3d12 -FeatureLevelEs31; older DX11 Performance instructions can select a different path. Epic competitive settings; Epic renderer persistence.

Riot added AMD Anti-Lag 2 in patch 12.09 for compatible hardware and sufficiently recent drivers. It is game integration, distinct from regular driver Anti-Lag. NVIDIA Reflex/Boost cannot be enabled on AMD. Riot 12.09; AMD driver controls; NVIDIA Reflex SDK availability.

Registry and service decisions

Apply Windows profile enables Game Mode and disables Windows background recording/controller Game Bar prompts using the existing backed-up preferences. Registry/Latency-Profile.reg is the same subset for review, not an additional installation step. Use the app for journaled changes.

No new system-service kill list is applied in response to a subjective delay report. Existing optional reductions remain, while HID/USB/controller, audio, network, graphics, capture, Windows servicing and anti-cheat dependencies stay available. A count of mostly idle service hosts does not establish how much work competes with the game.

Compact hosting is off by default on new 0.8.0 installations. It remains an opt-in count experiment. Existing installed flags remain until Restore host isolation is used and Windows restarts. That provides a useful comparison; there is no evidence yet that compaction caused this regression. The earlier 143-to-120 count was user-reported for 0.3.2. No new 60–70 count or measured latency result is claimed.

HAGS is also opt-in during installation. Existing values are not reset by leaving the installer option unchecked. Compare On and Off with the actual driver and games, restarting between choices. Microsoft's explanation describes a scheduler change, rather than a universally fastest setting. Microsoft HAGS.

For compatible DX10/11 games in windowed/borderless mode, Windows' Optimizations for windowed games can reduce presentation latency through the flip model. Fortnite's current DX12 Performance renderer uses a different path. The app opens the supported Windows graphics page; it does not claim a private compatibility registry override fixes every game. Microsoft windowed-game optimization.

The MMCSS facts are from Microsoft's scheduler reference, and timer guidance from Microsoft BCDEdit /set. Other omissions are engineering decisions under the available evidence, not results of a new benchmark.

Test your 9800X3D / RTX 5070 system

1. Record the GPU driver version and monitor's actual active refresh. For a controller, compare direct wired USB with the current connection without changing sensitivity or controller settings.
2. Apply Fortnite Low/Far, select DX12 Performance and Reflex On + Boost, then launch normally with anti-cheat. Let a repeatable practice/replay scene settle after game/driver/renderer updates. Do not clear shader caches between runs.
3. Capture five comparable runs of the same scene. Record average FPS, frame-time percentiles/1% lows and any available PC latency statistic. Keep resolution, controller, overlays and capture path constant.
4. Compare Reflex On versus On + Boost, then uncapped versus a sustainable cap. A cap can trade peak FPS for consistency on a saturated or thermally limited system; that is a hypothesis to test here.
5. If compact hosting is active, restore host isolation and restart for a separate comparison. Test HAGS On/Off with restarts and Gaming/Balanced/Max Performance power separately. Watch temperature and sustained clocks; the 100% AC CPU policy is a request, not a fixed-frequency or faster-FPS guarantee.
6. Test required OBS/capture/Voicemeeter workloads in their normal configuration as well as inactive. Keep their drivers and required services intact. Compare optional overlays individually.
7. If delay persists, choose Save latency configuration report. Pair that local JSON with Video-menu screenshots and a measurement capture. The report does not measure delay itself.

FrameView can report PC-latency metrics in instrumented games; Fortnite may require Latency Markers, and unavailable readings may be N/A. PC latency omits parts of the peripheral/display path. For a click-to-photon claim, use suitable Reflex Analyzer/LDAT equipment or a carefully designed high-speed-camera test. NVIDIA FrameView guide; Reflex Analyzer setup.

Networking and power research added to this release

Input-to-display latency and network round-trip latency are different measurements. Renderer/Reflex changes concern the former; local NIC processing and the internet route concern the latter. Microsoft describes packet batching through interrupt moderation and a tradeoff against CPU/interrupt work. Disabling moderation is an explicit test, not a universal gaming default. RSS normally distributes receive processing; the app only enables a supported standardized driver property and preserves other offloads. These do not remove game-server distance or router queues. Read network research and setup.

The new Max Performance scheme uses documented power settings. Microsoft defines minimum processor performance as a percentage request; the setting does not promise actual clock frequency on a CPPC/boost-managed Ryzen CPU. AC minimum/maximum 100 and EPP 0 are a high-power comparison, with PCIe/USB saving disabled where supported. The plan is created but not automatically selected at install. Compare it with Balanced and the older idle-scaling Gaming plan, with temperature and frame-time capture. OEM/Modern Standby restrictions may prevent additional plans; the app reports this rather than changing firmware or sleep capabilities.

Primary references: Microsoft powercfg, minimum processor performance, AC readback API, interrupt moderation.

Honest application status

Orange means saved configuration matches, not that latency improved. A registry value can be pending a restart. A game can overwrite local values through cloud sync or a new format.

Performance/Reflex/Anti-Lag 2 are selected once in game. Save in-game choices captures supported existing graphics entries for later replay; private enum meanings are not guessed. Some options may remain exclusively in game/cloud. NVIDIA profiles are per application; the AMD ADLX base is explicitly global and may be overridden by Adrenalin application profiles. See Game profiles.

This release has Linux-host parser, pure-logic, mock-driver, C# compilation, recovery and archive checks. Windows rendering/UAC, actual vendor API calls, game/cloud behavior, anti-cheat compatibility and gaming performance still require a Windows test. No absolute minimum delay, continuous process-count ceiling or compatibility with every future update is guaranteed.