Releases · JamesCJ60/Universal-x86-Tuning-Utility

01 Feb 18:57

JamesCJ60

2.5.7

ae21cb3

Universal x86 Tuning Utility V26.1.0 - Beta 0 Pre-release

Pre-release

Changes Made:

Removed all legacy WinRing0 and InpOutx64 driver dependencies and binaries from the project (Intel support has been removed temporarily due to this).
Introduced a new PawnIO-based backend for SMU access on AMD platforms - You will now need to download and install the PawnIO driver!
Updated .NET target to net10.0-windows10.0.22621.0
Various bug fixes and improvements!

Software Prerequisites:

Downloads:

Windows

Portable Download

Assets 4

17 Dec 20:44

JamesCJ60

2.5.6

ae21cb3

Universal x86 Tuning Utility V2.5.6 - Stable Latest

Latest

Changes made:

Includes changes made by @DenisGZM (Thank you!)
Various bug fixes and improvements!

This will be one of the last stable releases to utilise WinRing0. The upcoming beta releases will aim to migrate the vast majority of functionality to PawnIO.

Downloads:

Windows

UXTU ML Adaptive Mode Test 0.0.5 (Separate Application):

The purpose of this ML-based tool is to automatically tweak the CPU’s/APU’s/SmartShift power limit as you play. The CPU/APU/SmartShift algorithm allocates more power when the game becomes demanding and frame rates drop. If performance is well above your target, it reduces power draw to help keep temperatures and battery use in check while maintaining optimal performance.

The idea is based on how modern consoles, such as the Nintendo Switch and PS5, have predefined power profiles for specific performance ranges. This approach is enhanced by dynamically adapting in real-time using machine learning to tailor the performance to the device's characteristics for each game.

Advanced features/functionality in use:

Uses a lightweight, real-time reinforcement learning algorithm to decide how to adjust power limits for best gaming performance and efficiency.
Continuously monitors actual FPS, CPU/GPU usage, and temperatures, learning from real-world results instead of relying on preset rules.
Tracks the effects of its power changes over time, so it learns what works (and what doesn’t) in different gaming scenarios.
Dynamically balances between trying new power strategies and sticking with what it has already learned, based on how steady game performance is.
Assigns rewards or penalties to past decisions based on long-term outcomes, not just instant results, which helps avoid unstable or jittery behaviour.
Detects sudden changes in workload (like a new scene or a boss fight) and quickly adapts its strategy to suit new conditions.
Responds aggressively if FPS drops well below the target, increasing power more quickly to recover smooth performance.
Reduces power more aggressively (but not excessively) if FPS is far above the target, saving battery and reducing heat.
Remembers average power preferences for each game, so it can start from an optimal point next time you play.

Framework Laptop 16 Demo:

Downloads:

Windows

Portable Download

Contributors

DenisGZM

Assets 4

28 Sep 10:15

JamesCJ60

2.5.5

a618c18

Universal x86 Tuning Utility V2.5.5 - Stable

Changes made:

Updated premade presets for newer AMD APUs/CPUs and Framework Laptops

Reminder: The AMD and Intel backend for this application will be updated in the future, once 0Ring has been released and compatibility has been tested. This will address Windows Defender detecting Winring0 and the installer as a vulnerable driver/application.

Downloads:

Windows

Installer Download

Portable Download

UXTU ML Adaptive Mode Test 0.0.5 (Separate Application):

Advanced features/functionality in use:

Uses a lightweight, real-time reinforcement learning algorithm to decide how to adjust power limits for best gaming performance and efficiency.
Continuously monitors actual FPS, CPU/GPU usage, and temperatures, learning from real-world results instead of relying on preset rules.
Tracks the effects of its power changes over time, so it learns what works (and what doesn’t) in different gaming scenarios.
Dynamically balances between trying new power strategies and sticking with what it has already learned, based on how steady game performance is.
Assigns rewards or penalties to past decisions based on long-term outcomes, not just instant results, which helps avoid unstable or jittery behaviour.
Detects sudden changes in workload (like a new scene or a boss fight) and quickly adapts its strategy to suit new conditions.
Responds aggressively if FPS drops well below the target, increasing power more quickly to recover smooth performance.
Reduces power more aggressively (but not excessively) if FPS is far above the target, saving battery and reducing heat.
Remembers average power preferences for each game, so it can start from an optimal point next time you play.