The application

It shows its working.
All 386 of them.

Not a black box with a big red OPTIMISE button. Every tweak is listed, explained, individually toggleable, and individually reversible.

Forge Tweaks benchmark screen, showing the score, the saved before, and a per-component breakdown.
  • Higher 1% lows

    Stutter is not low average FPS — it is the floor collapsing when the fight starts. That is what we go after first.

  • Lower input latency

    Shorter frame queues, no acceleration you never asked for, and a display pipeline that stops holding frames back.

  • Nothing you can't undo

    Every value is read before it is written. Revert restores yours — not a default someone guessed at.

Pricing

Pay monthly.
Or once, and keep it for life.

  • Free

    Measure your machine and see where it actually stands.

    $0

    no card, no account

    • Full benchmark — CPU, memory, disk, latency
    • Your before, saved locally
    • Eight safe tweaks, fully reversible
    • The full tuning pass
    • Access to all profiles
    • Support
    Download free
  • Standard

    The whole tuning pass, with the before and after to prove it.

    $15

    USD · yours for life

    • Everything in Free
    • Every change in the table above
    • Before / after report
    • One-click full revert
    • Free updates forever
    • Access to all profiles
    Get Forge Tweaks
  • Most bought

    Premium

    Adds per-game tuning and a person who answers.

    $25

    USD · yours for life

    • Everything in Standard
    • Access to all profiles
    • Config-file locks so games stop undoing your settings
    • Priority support from a human
    • Access to everything
    Get Premium
  • Early access

    FrameForge

    Frame generation for the games that never shipped with it.

    $17

    USD · yours for life

    • Doubles the frame rate in FiveM and GTA V
    • Hardware optical flow on NVIDIA, software fallback elsewhere
    • Output pacing — even delivery, not just a doubled counter
    • Hotkey toggle, on and off mid-game
    • Never lowers your native frame rate
    • Anything in Forge Tweaks
    Get FrameForge

All prices in USD. Licensed per PC — change your GPU, CPU or drives and it keeps working, and we reissue free after a Windows reinstall. Cancel a monthly plan whenever you like; a lifetime purchase never expires.

FrameForge

Double the frames.
From the game you already own.

FiveM and GTA V are DX11. There is no DLSS-G or FSR3 path for them, so they have no frame generation at all — that gap is what FrameForge fills. Drag the divider: same clip, same moment, left half getting half the frames.

Native

0fps

FrameForge

0fps

Drag the divider

Native

Forged

Frames delivered · last second

FiveM · 120 fps source

FrameForge sits between your game and your screen, measures how every part of the picture moved between the last two frames it drew, and builds the frame that belongs in between them. It then holds that frame until the moment it is due, so the extra frames arrive evenly instead of in pairs — which is the difference between a doubled counter and motion that actually feels doubled.

  1. 01

    Read the two real frames

    The frame the game just drew, and the one before it. Nothing is guessed about what the game is doing — both inputs are finished frames.

  2. 02

    Work out what moved

    Optical flow measures where every part of the image travelled between those two frames. On NVIDIA that runs on the dedicated optical-flow hardware.

  3. 03

    Build the frame in between

    Each pixel is carried half-way along the motion it was found to have, forwards from one frame and backwards from the other, then blended.

  4. 04

    Pace the output

    The extra frame is held until the moment it belongs, rather than shown the instant it is ready. This is the difference between a doubled counter and motion that actually feels doubled.

Measured

Interpolation cost per frame
1.55 ms
Hardware optical flow. The software path costs 9.6 ms and is used when there is no NVOFA.
Headroom at that cost
~646 fps
What the interpolation stage alone could sustain, before the game's own frame time.
Effect on native frame rate
None
The hard constraint the whole design is held to: generated frames must never cost real ones.

Measured on the development machine — Ryzen 5 5600X, RTX 4060, 1080p — using the build's own timing harness.

Get FrameForgeEarly access · FiveM and GTA V

Forge Tweaks · four games, one machine

Milliseconds matter.
These are the medians.

Fortnite, FiveM, Valorant and Warzone, measured the same way each time on a Ryzen 5 5600X with an RTX 3060 Ti. The middle result of the four, not the best one — your gain depends on how much was wrong to begin with.

  • +46%

    1% lows

    Median improvement

  • −32%

    Input latency

    Median reduction

  • +15%

    Average FPS

    Median improvement

The process

How does it work?
Three steps, nine minutes.

Nothing runs unasked, and you see the change before you approve it.

  1. 01About 2 minutes

    Measure

    You
    Run the benchmark. No account, no sign-up.
    Forge
    We record CPU, memory, disk and latency, and save it as your before.
  2. 02About 5 minutes

    Tune

    You
    Review the list of changes and approve it. Nothing runs unasked.
    Forge
    Each change is applied, then read back to confirm it took. Restart point created first.
  3. 03About 2 minutes

    Prove

    You
    Restart, run the benchmark again, compare.
    Forge
    Both numbers shown side by side. If a change did nothing, we say so.

Full disclosure

Exactly what we change.
Published before you buy.

A page selling PC optimisation that won't tell you what it does to your machine until after payment has already answered your question.

None

GPU scheduling

Shortens the queue between your click and the frame that shows it.

To undo: Single registry value, restored exactly as found.

Worth knowing

Power plan

Stops the CPU dropping clocks in the half-second a fight starts.

To undo: Original scheme reactivated; your old plan is never deleted.

Low

Network stack

Cuts the delay between your input reaching the server and the reply.

To undo: Every value snapshotted before the change and written back on revert.

Worth knowing

Background services

Reclaims CPU time and disk I/O that spikes exactly when you're loading in.

To undo: Start type restored to its original value, not to a guessed default.

None

Visual effects

Frees a small amount of GPU time and makes the desktop feel instant.

To undo: Restored with one click.

None

Storage

Recovers space and stops stale caches slowing load times.

To undo: Nothing to undo — this only deletes files Windows regenerates.

What we never touch

  • Windows Update — you keep getting security patches
  • Windows Defender — never disabled, never excluded
  • Drivers — we don't install, replace or 'optimise' them
  • The kernel — no drivers of our own, no ring-0 anything
  • Your files — nothing outside Windows settings is touched
  • Anti-cheat — nothing we change hooks or hides from it

Safety

Safe, reversible,
and yours.

Four things that are true of this and are not true of most of what gets sold in this category.

  • No remote access, ever

    We never connect to your machine. There's no TeamViewer link, no AnyDesk code, no screen share. Everything runs locally and works offline.

  • No kernel drivers

    Nothing we install runs at ring-0. That's the layer anti-cheat cares about, and it's the layer we deliberately stay out of.

  • Every change snapshotted

    Before a value is written, the original is read and stored. Revert restores your value — not a default someone guessed at.

  • Restore point first

    A Windows restore point is created before the first change of every session, so there's a floor underneath everything.

Guarantee

7 days, money back, no argument

Run the benchmark before and after. If the numbers didn't move on your machine, email us and we refund you. We don't ask you to prove anything — we just take your word for it, because the benchmark already made the case or it didn't.

Read the terms

Support

Frequently asked
questions.

Will this get me banned by anti-cheat?

No. Nothing we change touches game memory, injects into a game process, or installs a kernel driver — those are the three things anti-cheat actually looks for. Every change is a Windows setting: registry values, service start types, power plan entries, network stack parameters. Vanguard, EAC, BattlEye and FACEIT are all indifferent to those. If you want to check, the full list of changes is on this page before you buy, not hidden behind a purchase.

Is this a scam?

Fair question — most things in this category are. The difference is that we measure your machine before and after and show you both numbers, so you don't have to take a claim on faith. If the number didn't move, the report says it didn't move, and there's a 7-day refund. We also publish exactly what gets changed, which is the thing a scam can never afford to do.

Will it break my PC?

It shouldn't, and there are three layers underneath you if something goes wrong. A Windows restore point is created before the first change. Every value is read and stored before it's overwritten, so revert restores your original rather than a guessed default. And nothing we touch is outside Windows' own settings — no drivers, no firmware, no files of yours.

Can I undo it?

Yes, all of it, with one button. This matters more than it sounds: a lot of 'optimiser' tools apply changes they cannot reverse, because they never recorded what was there before. We snapshot every value before writing it. Two things are one-way by nature — deleted temp files and a Windows restore point you've since removed — and both are noted at the time.

Do you need remote access to my machine?

Never. There's no TeamViewer session, no AnyDesk code, no screen share, and no account. Everything runs locally on your PC and works with the network unplugged. If anyone selling PC optimisation asks for remote access, that alone is a reason to walk away.

How much FPS will I actually gain?

Honestly: it depends entirely on how much was wrong to begin with. A freshly installed Windows on good hardware might move 5%. A four-year-old system with a dozen startup apps, recording overlays and a balanced power plan can move 30–40% on 1% lows. The benchmark tells you which one you are before you spend anything, and the average gain is far smaller than the headline gain.

Does it work on my hardware?

If you're running Windows 10 or 11 on an Intel or AMD CPU with an NVIDIA, AMD or Intel GPU, yes. Some changes are vendor-specific and are skipped automatically on hardware they don't apply to — the report shows you which ones were skipped and why, rather than pretending it applied them.

Why not just use a free script from YouTube?

You can, and some of them are fine. The problems are that they rarely measure anything, they usually can't be undone, and a meaningful number of them disable Windows Update or Defender to make a benchmark look better — which is a genuinely bad trade you weren't told you were making. What you're paying for here is the measurement and the revert, not secret knowledge.

Do I need to keep it installed?

No. The changes are made to Windows itself, so they persist whether the app is running or not. Keep it if you want the revert switch and the benchmark handy — plenty of people run it once and uninstall.

Still have questions?

Email a human ↗

Measure it. Then decide.

The benchmark is free and takes two minutes. If your machine is already fine, it will tell you that.