Processor scheduling · a simple Fortnite comparison guide

LeCoreOS 11 0.8.5 · cooked by LeSamsky · reviewed 17 October 2026

My recommendation

Start with Programs — 0x26 hexadecimal / 38 decimal. This follows the traditional Windows client foreground-oriented scheduling policy. It is my starting recommendation, not a measured winner on your PC. If your current setting already feels good, keep it. Installing this app update does not change it.

The supplied image lists twelve registry combinations. I also read the supplied Medium article, published in 2017. It describes older Windows scheduling behavior; it does not provide modern Fortnite input-to-display tests. None of the primary sources below establishes one of these twelve values as the fastest for every Windows 11 build and CPU.

What the words mean

CPU turn / quantum: how long a runnable thread can continue before another equal-priority thread gets a turn. A thread can wait, yield or be preempted earlier. This is not a fixed input delay.
Short / long: selects a time-slice policy. It is not a promise of a particular number of milliseconds on Windows 11.
Variable: the foreground field can extend CPU turns for the focused application. It does not randomly vary FPS.
Fixed: uses a fixed quantum table. The chart's high/medium field labels do not create the variable table's 3:1/2:1 quantum extension. Do not assume 0x2A is a stronger or faster 0x28.
Foreground: the application you are actively using. This system-wide setting is not a Fortnite-only profile, GPU setting, timer-resolution control, controller polling rate or network-ping control.

Microsoft's current CPU-analysis documentation describes short client quanta with foreground extensions and immediate preemption by higher-priority ready threads. The detailed bit-field explanation comes from older Windows Internals material, not a current Windows 11 Fortnite benchmark. Dynamic input and I/O priority boosts are separate mechanisms.

Pick and apply in the app

Open Control Center → Gaming → Processor scheduling. Each option has a short visible explanation above its Apply button. Programs and Short Fixed appear first; expand Advanced comparisons for all other chart values.

Hexadecimal	Decimal	Slice policy	Foreground field
0x26	38	Short, variable — recommended starting point	2; larger extension
0x28	40	Short, fixed — first optional comparison	0
0x25	37	Short, variable	1; smaller extension
0x24	36	Short, variable	0; no extension
0x2A	42	Short, fixed	2; not a variable 3:1 extension
0x29	41	Short, fixed	1; not a variable 2:1 extension
0x16	22	Long, variable	2
0x15	21	Long, variable	1
0x14	20	Long, variable	0
0x18	24	Long, fixed; background-throughput style	0
0x1A	26	Long, fixed	2; fixed table
0x19	25	Long, fixed	1; fixed table

Decimal 26 is hexadecimal 0x1A, not 0x26. The app labels both to prevent this common mistake. It writes only a selected allowlisted DWORD at HKLM\SYSTEM\CurrentControlSet\Control\PriorityControl\Win32PrioritySeparation. An existing value such as 0x02 is shown as Windows client defaults; it is not silently rewritten or falsely highlighted as 0x26.

Close games, click the chosen Apply, accept normal administrator elevation using the original gaming account, and restart Windows. Orange means the stored DWORD matches. It does not prove that the running kernel has adopted the setting or that latency improved. Use Save configuration report to record the configured value; that report does not measure latency.

Restore and backups

Before a different value is written, the app exports the PriorityControl key and saves the original value, type and presence in the protected %ProgramData%\LeGameOS\ProcessorScheduling.json journal. Exports are in LatencyRegistryBackups\PriorityControl-*.reg.

Restore original scheduling returns to the value from before your first successful experiment in that comparison cycle. Switching between presets does not replace it. An originally absent value is removed; an original non-DWORD type is preserved. The app checks writes, rolls a failed change back when it can, and retains an unfinished recovery record if rollback fails. A later outside edit is kept with a warning. Do not delete the journals or import a whole exported key without reviewing it.

Processor scheduling Restore is separate from Restore MMCSS and Restore core parking. Machine recovery and Restore all latency comparisons also attempt processor scheduling recovery. Restart after restoration.

Find the smoothest choice on your own PC

Save a configuration report and keep your existing good setup as the comparison baseline. Use the gaming display connected directly to your GPU. Keep your streaming/capture route consistent; compare with gaming-PC OBS projector/preview closed where possible.
Keep Fortnite resolution, rendering mode, graphics, Reflex, FPS cap, drivers and background apps identical. Warm up the same Creative island/scene first, especially after a game or driver update.
Test A: Programs 0x26, restart, then make three comparable runs. Record frame-time variation, 1% lows and PC-latency readings if your measurement tool supports them.
Test B: Short Fixed 0x28, restart and repeat. Then return to A and repeat to check whether an apparent change was normal run-to-run variation. Do not change MMCSS, HAGS, parking and scheduling together.
Keep a change only when its improvement repeats without worse frame pacing, audio or capture behavior. If results are equal, keep Programs or restore your original. Long/fixed alternatives are optional experiments, not automatic upgrades.

Average FPS alone does not measure input-to-display delay. FrameView needs supported latency markers to report PC latency; PC latency is not the complete controller-to-screen path. End-to-end comparisons need suitable measurement hardware or a carefully controlled high-speed camera method. A configuration report cannot manufacture those measurements.

Fortnite settings with stronger direct guidance

Epic's April 2026 competitive guide recommends Performance rendering, low detail, VSync Off and NVIDIA Reflex On + Boost. Your existing Low / Far preset stays unchanged; Epic suggests Near view distance, which you can compare manually if you prefer. Keep the same choice in both scheduling tests. Reflex manages render submission; a registry time-slice change does not replace it.

Your 9800X3D / RTX 5070 system should be compared with its current chipset/GPU/controller drivers and the monitor's highest supported refresh rate. Keep controller and audio dependencies, including your required routing software. Do not judge a scheduling value by process count, network ping or how aggressive its name sounds. There is no zero-delay preset.

Sources and limits

Microsoft: CPU Analysis — scheduling, quantum, preemption and CPU contention.
Microsoft: Priority Boosts — dynamic foreground/input/I/O boosts.
Microsoft Press: Windows Internals, Sixth Edition, Part 1 sample chapters — historical quantum tables/registry fields, chapter 5, pages 426–428. Exact modern timing and game gains are not established by this source.
Epic: Fortnite on PC — competitive settings in 2026 — supported in-game recommendations.
NVIDIA: System Latency Optimization Guide — Reflex and the rendering queue; originally published in 2020, so older presentation-mode advice is not a universal statement about current Windows.
NVIDIA: FrameView 1.4 User Guide — Fortnite latency markers and the meaning of recorded metrics.

Supplied reference reviewed: 2017 Medium article, used to identify the requested chart options. Performance conclusions above rely on primary sources and are marked as recommendations or measurements still required.

This release is tested with automated parsing, mocked registry/recovery cases, UI structure and archive checks. Native Windows WPF checks run during the update on your PC. AME installation, reboot behavior and actual Fortnite latency on your hardware have not been measured here.