Pavise: a Windows game resource manager that undoes what it changes

The contention problem Pavise was built around

A game on a Windows desktop rarely runs alone. Browsers, launchers, update services and sync clients keep working while the game holds the foreground, and they compete for CPU time, disk I/O, paging and GPU attention. Pavise's stated job is to manage that competition: the README says it "gives the game more scheduling priority and reduces the resources claimed by eligible background processes while you play." That is a narrower claim than most optimizer tools make, and the README repeats the limit explicitly: "It does not create extra CPU or GPU performance." The intended user is someone who has identified real background contention on their own machine, not someone shopping for a frame-rate number. The project ships in English, Simplified Chinese and Japanese documentation, and the interface supports English and Simplified Chinese.

How the session mechanism works, and why recovery is the hard part

The design is a session with a recorded before-state. When you enable the guard and launch a detected game, Pavise applies policies: process priorities across CPU, disk I/O, paging and GPU, EcoQoS and timer policy, optional core assignment, and optionally driver-level graphics settings. Each change is recorded with its original value. When the game exits, the README states that Pavise "automatically restores the session changes it recorded." If Pavise itself exits unexpectedly, the next launch retries recovery and keeps records of anything it could not restore. That recovery path is the most interesting engineering decision here, and also the least documented: the README points to a "What gets restored" section for the different rules applied to persistent settings and file cleanup, but the excerpt available does not spell those rules out. The project also states it runs locally, uploads no local machine data, and does not inject into game processes or modify game memory. No injection is a meaningful boundary for anyone nervous about anti-cheat, though it also means Pavise can only influence scheduling and device policy, not the game's own threads.

Installing Pavise and running a first guarded session

There is no package manager step. The README directs readers to the official download page for the release notes and a choice between GitHub and Quark mirrors, and states that downloads are free. Requirements are Windows 10 version 2004 (build 19041) or later, with Windows 11 24H2 preferred; resource management needs administrator privileges. After installing, the first real use is adding a game and starting a session. The README's quick start describes adding an EXE, shortcut or folder, or scanning installed libraries from Steam, Epic, GOG, Ubisoft, Riot, WeGame, Battle.net, Xbox and Microsoft Store. The repository also contains a rescue script at the top level, Pavise-Rescue.cmd, which is worth knowing about before you enable anything, since it is the file you would reach for if a session ends badly.

The flow below mirrors the quick start steps, with the mode selection and whitelist review that the README asks for before play.

What the four modes actually change

Mode selection is the main decision a new user makes, and the table in the README separates them by background scope rather than by performance. Smart isolates eligible background processes when the session starts and exempts the app you are using and its family after an alt-tab. Esports widens suppression to eligible non-game processes, including windowed apps, so the app you alt-tab into stays eligible; whitelist and built-in protection still apply. Handheld uses the Esports background scope but leaves power control to vendor tools, requires a battery, and excludes several desktop-oriented policies. Custom lets you pick background, core, graphics, memory, power and environment policies individually. Modes your machine cannot use are hidden, so a desktop will not show Handheld. The former Extreme mode was removed in v2.2.2; its additional controls became separate switches, off by default. That removal is a reasonable simplification, but it also means anyone following an older guide will look for a mode that no longer exists.

Protection rules, anti-cheat and the limits of suppression

Suppression is not indiscriminate. Ordinary background suppression exempts anti-cheat processes, Windows core services, the input, audio and peripheral chain, hardware tools and other signed-in accounts. Game-family exemption is on by default. Separate anti-cheat controls exist, are off by default, and cover selected user-mode processes only; the README is explicit that they do not control kernel drivers. That sentence matters more than any feature list. If a game's anti-cheat runs in kernel mode, Pavise's anti-cheat switches cannot touch it, and the practical question of whether priority changes are tolerated at all is one the documentation does not answer. Treat any anti-cheat title as a case to test deliberately, with a session report to check afterwards, rather than a default. The whitelist is your other control surface, and the README asks you to add applications you want kept unaffected before enabling the guard.

Where Pavise is the wrong tool

The README's own caveat is the clearest limitation: the benefit depends on background contention, hardware and the game, and "a clean system or a fully GPU-bound game may see little change." If your frame times are limited by the GPU, no amount of process priority will move them. Two other cases stand out. First, machines without a battery cannot use Handheld mode, and users who already manage power through vendor tools are told Handheld leaves power control to those tools anyway. Second, anyone who wants a permanent, system-wide tuning setup will find the session model working against them: Pavise restores what it changed, and the README notes that persistent settings and file cleanup have different recovery rules, so a setting you expect to survive may not. The project also requires administrator privileges for resource management and checks its official update source at startup, which is a network call some locked-down environments will not permit.

How Pavise differs from Process Lasso and manual affinity tools

Process Lasso is the obvious comparison for anyone who has already tried fixing contention by hand: it applies persistent rules to processes, and those rules stay applied. Pavise inverts that. Its policies are scoped to a play session and are recorded so they can be reverted, and its per-game profiles exist so you do not have to re-tune your everyday desktop. If your goal is a machine that is always tuned a certain way, a persistent rules tool is the better fit. If your goal is a machine that behaves normally when you are not gaming, the session model is the point. Manual affinity and priority changes through Task Manager sit at the other extreme: no profiles, no recording, no recovery, and no protection rules for anti-cheat or the input chain. Pavise's value is the bookkeeping around the changes, not the changes themselves.

Licence, maintenance and upgrade cost

Pavise is licensed GPL-3.0, and the repository carries a NOTICE file alongside the LICENSE. For end users that changes little; for anyone embedding Pavise in a commercial product or shipping a modified build, the copyleft terms apply to distributed derivatives, and reading LICENSE and NOTICE is the place to start rather than a summary. Maintenance signals are current: the last push to the default branch pavise2x was on 2026-09-15, and releases v2.2.2.3, v2.2.2.2 and v2.2.1.3 landed on 2026-09-14, 2026-09-11 and 2026-09-07 respectively. That release cadence also sets the upgrade cost: this is a fast-moving utility, and the removal of Extreme mode in v2.2.2 shows that settings you configure today can be restructured. Expect to re-read release notes after updates rather than assuming your profile survives untouched.

Editorial conclusion

Pavise is for Windows players on build 19041 or later who can name the background processes stealing time from a specific game and want those changes reverted afterwards. It is not for anyone expecting raw FPS gains, and not for handhelds that lack a battery or rely on vendor power tools. Before adopting it, verify that your anti-cheat title tolerates priority changes at all, since Pavise's separate anti-cheat controls are off by default and cover only selected user-mode processes; then check the session report after one play session to confirm every recorded change was restored.

Frequently asked questions

Official sources

1. dulaiduwang003/Pavise-Game on GitHub
2. License: GPL-3.0
3. Project website
4. README
5. Releases