Do game boosters and optimizers actually work?

Sometimes, and by less than they claim. A game booster or PC optimizer can only change what Windows lets any program change: settings, services, scheduled tasks, the priority of running programs. It cannot make the graphics card faster, and in a game that waits on the graphics card, the card sets the frame rate. Some of these tools leave a PC slower or less stable than before, and a "50%" printed for every PC was never measured on yours.

The rest of this page is the long answer: what such a tool can and cannot change, the kinds of tool sold under these names, and how to check one before you trust it. It names no product. It is written by the makers of one of these tools, so read it as that, and check what you can.

What any optimizer can change, and what none can

A program running with administrator rights can change:

• Windows settings: registry values, power plans and processor power settings, DNS servers, visual effects, notifications.
• Background work: which services start and how, scheduled tasks, startup entries, the search indexer.
• How Windows schedules a running program: its priority, and the processor cores it may use.
• The settings a driver offers to programs, such as a graphics driver's profile for one game.
• Files it is allowed to write, a game's own config files among them.

No program can change:

• The hardware: how fast the graphics card renders, the processor's power limit, the memory speed the firmware set, the refresh rates the monitor supports.
• The game's code: how the engine builds a frame, how often it reads input, how often its servers update.
• The network past your router: the route from your provider to the game server, and how busy that route is.

Whatever a tool claims has to fit inside the first list. The same split holds for input lag: of the dozen stages between a click and a changed pixel, software owns three, and The input lag no software can fix names the owner of each.

Why the same tool helps one PC and does nothing on another

Everything in the first list is a setting, and a setting costs a game something only when it is wrong for that game. So the most any optimizer can do is take back a cost that something else was already imposing: a power plan that lets the processor drop its clock in the middle of a fight, a scheduled task that wakes during a match, a game left on the integrated graphics, a monitor still at 60 Hz after a cable swap, a background program on the cores the game wanted.

On a PC where those were already right, there is nothing to take back, and the same tool changes nothing you can measure. That is why two honest people can report opposite results from one program: their machines started in different states. It is also why a fixed percentage on a download page cannot be true for everyone who reads it. Windows 11 gaming settings that actually matter goes through the settings worth checking by hand, and who each one helps.

The kinds of tool sold under these names

Five kinds, often several of them in one program. Each is useful in a narrow case, and each has its own way of going wrong.

Priority and core-affinity managers

They raise the game's priority, lower everything else, or pin programs to chosen cores. That helps when programs really compete for the processor: a background job, a stream encoder, an update running during a match. When cores sit idle, priority decides nothing, because there is no queue to jump; Task Manager for gamers shows how to see whether anything competes at all. Pinning a game to fewer cores than it uses can make it worse, and on processors whose cores differ the wrong pin puts the game exactly where it should not run: Intel P-cores and E-cores in games and Ryzen X3D with two CCDs cover those two cases.

Registry tweak packs

Lists of registry values applied in one go. A few do what they say; many date from older versions of Windows and do nothing now; some do harm. w0a refuses two groups for that reason: the kernel-level memory values, and the display-driver keys that switch off multi-plane overlay or change how Windows recovers a hung graphics card. On a real machine those display keys froze CS2. A pack applied in one go also leaves no way to tell which value did what.

Debloat scripts

Scripts that remove built-in apps, services and scheduled tasks in bulk. Removing what you never use frees little, because a service that is not running costs almost nothing. The risk is what goes with it: a script that takes out Windows Update, the Store or a security component leaves a system that cannot update or reinstall what it later needs, and a script that keeps no record of what it removed leaves nothing to put things back from.

Network route services

Services that carry your game traffic through their own servers. A different route can lower ping when your provider's route to the game server is poor, and it cannot help when that route was already good, because the extra stop adds its own distance. No setting on your PC changes the route past your router, so a tool that only changes settings cannot do what a route service does, and a route service cannot fix a crowded Wi-Fi channel in your home. Can a VPN lower your ping? shows how to test whether another route helps you.

Modified Windows builds

Copies of Windows rebuilt for gaming, with components removed before installation. They carry a debloat script's risks for the whole system, and their changes cannot be listed or undone: short of reinstalling, there is no unmodified Windows to go back to. Updates can fail or be held back, and an anti-cheat that checks Secure Boot, TPM or memory integrity can refuse to start on a build that turned them off (Secure Boot and TPM 2.0 for anti-cheat covers what it checks). A build downloaded from a stranger is also code you cannot inspect, running with every right on your machine.

How to check a tool before you trust it

Five questions. A tool that cannot answer the first two cannot be checked at all:

1. Does it list every change it makes, by name and value?
2. Does it record the old value, and can it put every change back?
3. Does it say what it refuses to do, and why?
4. Does it measure on your machine, before and after, rather than print one number for everyone?
5. Can you apply one change at a time, so you can tell which one did what?

A tool that does not list its changes and cannot undo them cannot be audited: neither you nor anyone you ask for help can tell afterwards what it did. If you already ran one, the Windows defaults are the way back, and on a machine where nobody knows what changed, writing the defaults back is the only repair that does not depend on that tool's memory.

What w0a changes, and what it does not touch

w0a is one of these tools, and the same five questions apply to it. It changes Windows settings: registry values, power plans, DNS servers, scheduled tasks and the start type of services on a named list. A tweak can only write a registry value on a path from a list compiled into the application; a write anywhere else is refused. Two kinds of switch work outside that list: the Startup page turns another program's startup entry on or off, and the in-game switches that hold notifications and update downloads back during a match put the value back when the game closes. A few pages change things of their own: CS2's launch options and video settings file and w0a's own config for the game, up to four documented settings in NVIDIA's profile for CS2, and the graphics card's clocks, power limit and fan speed, saved as profiles. Each has its own way back.

Not every change has a way back. The Uninstall Apps page removes a program you choose, a preinstalled one included; with deep clean on, it also deletes the folders, registry keys and other traces that program left behind. The Logs page permanently deletes the logs and caches you select. The Drivers page installs a newer version of an app through winget when you press Update, and that list can include drivers and runtimes. Deleting a power plan or old USB device records is final too. Uninstalling and log cleaning warn you and ask for confirmation before they start; the Backups page keeps a record of what they did, but it cannot undo any of these.

What it does not touch:

• Kernel-level values: paging executive, large system cache, NTFS memory usage and timer resolution requests are refused, and so is every value under Session Manager\kernel.
• The display-driver keys for MPO and TDR (OverlayTestMode, DisableOverlays, TdrDelay, TdrLevel): never written.
• Your games: no code is injected into them and none of their memory is read.
• A driver of its own: w0a ships none. For three hardware readings it can install a separately signed driver that only reads, and only after you press its button and confirm.
• Microsoft Defender: none of its settings are changed and no exclusions are added.

Nothing changes until you press a button or switch an engine on, there is no button that optimizes the whole PC at once, and every engine is off on a new installation. Before a tweak writes a new value, it records the one that was there; a tweak that only puts a Windows default back keeps no record of what it replaced, and neither does the Startup page, where switching the entry back is the undo. Pages that change Windows settings in a group write a checkpoint first; the Backups page lists them, and restoring one undoes everything applied after it. Nearly every card on the Fix page writes the Windows default back, so it also repairs a setting another tool changed; those cards keep no record of the value they replace. Undoing what w0a changed walks through both.

The engines, the parts of the Game Mode switch on the Process Manager page, are the first kind of tool above: they raise the game's priority, balance background work and can reserve cores for it. Whether they change frame times was not measured. What w0a's engines cost while a game runs has the one measurement there is, their processor cost on one PC, and how to run the comparison on yours.

Measure it on your own PC

The only result that applies to your machine is one measured on it:

1. Take a capture before you change anything: same map, same route, same length. w0a's Benchmark does this, and so does any frame-time tool.
2. Change one thing.
3. Take the same capture again, and compare the 1% low as well as the average.
4. Run the unchanged setup more than once. If those runs differ by more than your change did, the change is within noise.
5. Undo what did not move a number.

Reading the benchmark explains the figures and how to run an A/B series.

Related

• How to reduce input lag: what software can and cannot shorten
• How to optimize Windows 11 for gaming: the settings that actually matter
• What w0a's engines cost while a game runs
• Undoing what w0a changed
• Reading the benchmark