Competition options in LeCoreOS 11 0.8.0

Latency takes priority over a process-count target. Competition reductions remain selected. Experimental compact hosts are now unchecked by default; HAGS is also opt-in. Existing installations retain their previous values until restored. Use Restore host isolation and restart for a comparison if needed. No evidence establishes compact hosting as the cause of your Fortnite delay.

Retained background-reduction profile

Your measured baseline matters: you reported 143 idle processes before LeCoreOS 11 and 120 after the successful 0.3.2 installation. These are your observations, not measurements made here. Your latest Fortnite test reports noticeable delay; no instrumented timing was supplied. This release targets 60–70 at idle and retains two separate options: Competition background reductions and experimental compact hosts. The result on your PC still needs a restart and measurement.

What the new options do

Area	Applied change	Practical effect
Browser preload	Edge startup boost and background browser mode off	Closed Edge sessions should stop preloading; Edge/WebView2 remain installed.
Search	Search highlights policy off on supported editions	Removes online highlights; basic Start/app search remains.
Machine startup	Trim only the existing exact app-name patterns from the 64-bit and WOW6432Node Run keys	Apps remain installed and launch normally; device, voice and anti-cheat entries are excluded.
Scheduled tasks	Disable six named CEIP, Feedback/Siuf and offline Maps tasks	Prevents their future triggers; running work is not forcibly terminated. Updates/security/appraiser tasks remain.
Cross-device features	Disable CDPSvc and CDPUserSvc	Nearby sharing and connected-device handoff become unavailable. Controller driver and pairing services remain.
Clipboard	Disable cbdhsvc template	Targets clipboard history/cloud sync; verify ordinary copy/paste after restarting.
Other user services	Disable MessagingService and conditionally PrintWorkflowUserSvc	Windows phone/SMS integration removed; print workflow only reduced when Spooler is disabled.
Disabled template instances	Set supported UserServiceFlags to 0 after the template is already disabled	Prevents new instances on a subsequent sign-in; current instances are left alone.
Optional UI packages	Remove current-user Widgets, Game Bar and Xbox speech-overlay packages	Their UI/recording shortcuts go away. OBS/capture, Gaming Services and licensing are retained. Optional app recovery may need Store reinstallation.
Compact hosts	Request SvcHostSplitDisable=1 for eligible services from a fixed list of 48	Count experiment; services stay running, while eligible hosting can change after restarting.

Profiles/targets are readable in Profile.json; implementations are in LeGameOS.ps1 and Competition.ps1. The original gaming power, input, capture and Fortnite changes remain. The earlier permission fixes remain. An explicitly optional refused registry setting is warned and skipped; failed backup storage and unrelated registry errors still stop installation.

Compact hosts and performance

Windows service hosting is distinct from removing background work. Microsoft's service-host documentation describes the reliability/isolation benefits of separate hosts; that article describes Windows 10, not a 26H2 gaming benchmark. Grouping can reduce process boundaries and isolation. It does not establish faster game simulation, better frame times or lower input delay.

LeCoreOS 11 changes individual hosting flags; it does not force a giant global RAM threshold or rewrite service groups, executable paths, identities, privileges, startup types or service DLLs. Its count option accepts only running Win32 shared-process services from the explicit list, hosted by the system svchost.exe, with a System32 DLL and a valid Microsoft signature. Kernel drivers, own-process services, user templates, nonrunning services, untrusted DLLs and the isolated-role exclusions are skipped. All listed controller, audio, capture, security, anti-cheat and core exclusions stay out.

Eligibility depends on the installed Windows build and actual service configuration. The experiment is enabled for the four known build families 22631/26100/26200/26300 and skipped on unvalidated later builds. Signature or inventory failures skip the affected candidate rather than guessing. Each changed flag is journaled before writing, including original absence. Restart is needed; no live host is terminated or consolidated by the script.

Use Control Center 21 and restart to restore only the original hosting flags. Keep the Competition reductions if they work for you, and compare the same game with normal hosting versus compact hosting before keeping it. The new mode's service behavior and count effect remain untested on Windows.

Applying over your successful installation

Save work, close games and complete pending Windows updates.
Reopen AME Wizard and import LeCoreOS-11-0.8.0.apbx using the same gaming administrator account. Keep the existing user/machine journals.
Review the last page. Competition and compact hosts are selected for your request; either can be unchecked. Read the feature losses before applying.
Apply. A local process/service inventory is captured before the new changes. Existing Brave/Discord versions and profiles are kept.
Restart. With apps closed, wait about three minutes and select Control Center 19 for twelve idle samples. It reports min/median/max and splits service-host versus other processes without excluding anything.
Select 23 for process families, executable/parent IDs and service-to-PID mapping. Command lines/tokens are omitted. Reports stay local.

Neither the inventory nor the audit is a latency test. Both add their own temporary process while running; closing the tools removes that overhead. The sampled target is a maximum of 70 in that window, not a permanent cap. Updates, drivers, launchers and anti-cheat can create more processes later. There is no process-count enforcement agent and no essential-service killer.

Competitive game settings

The OS profile reduces background activity; the rendering pipeline determines much of the remaining PC latency. For Fortnite keep Performance rendering, low graphics, V-Sync off and frame generation off. For Valorant keep Multithreaded Rendering on and V-Sync off. In both, start by comparing NVIDIA Reflex On + Boost against On in the game. Use the highest display refresh rate and your existing controller/stretched/capture configuration.

NVIDIA recommends in-game Reflex for supported games; driver Low Latency Mode does not stack into a second Reflex benefit. Boost trades more GPU power for keeping clocks up in some CPU-bound workloads. Its result and the best FPS cap are workload-dependent. No driver profile blob or private game enum is injected here. The existing Control Center FPS-cap and HAGS comparisons remain available. See LATENCY-RESEARCH.html for a repeatable test plan.

The package does not introduce boot-timer, affinity, Realtime priority, blanket network-stack, pagefile-removal or security-disable tweaks. Changing every available registry value would not establish lower latency. For example, Microsoft documents that MMCSS SystemResponsiveness values below 10 clamp back to 20 and that its GPU Priority field is unused; such edits are omitted.

Wallpaper gallery and recovery

The desktop app's Wallpapers page (advanced console 24) offers six static 3840x2160 exports: Locker Room, Grand Library, Cosmic Court, Grand Prix Lounge, Carbon Apex and Neon Street Court. The third new attachment, Neon Street Court, is the default. Every image has the small Made by LeSamsky credit. Automatic sign-in rotation is On by default, keeps the default for the first sign-in and advances one scene on later sign-ins. Off removes the owned startup entry and holds the chosen image. The worker exits after applying the image; no resident wallpaper player runs. Read CONTROL-CENTER.html for startup details, desktop icon preservation and recovery.

Control Center 10 restores all saved settings, 21 restores only compact hosting, 8 restores printing/indexing including the new print workflow, and 11 handles optional app recovery. Task recovery checks recorded commands/principals and changes only the enabled state. It does not overwrite Windows-updated task definitions or recreate a deleted task. Missing/changed recovery targets are logged; keep the backups if anything remains unresolved.