AMD Device Library eXtra — ADLX SDK

ADLX is a modern SDK designed to access features and functionality of AMD systems across Display, 3D Graphics, Performance Monitoring, GPU Tuning, and more. It is targeted at developers and partners building control and monitoring applications for AMD devices.

📖 Full documentation: gpuopen.com/adlx

---

Key Features

• Modern API design — quality and reliability at the core
• Multi-language support — C, C++, C#, Java, Python
• Driver compatible — forward and backward compatible with AMD drivers
• Global state synchronization — consistent state across all callers
• Broad feature coverage — Display, 3D Graphics, Performance Monitoring, GPU Tuning, Power Tuning, and more
• AI/LLM integration — MCP servers for controlling AMD GPU settings via AI agents

---

Repository Structure

ADLX/
├── SDK/                  # Core ADLX interface headers and ADLXHelper wrappers (C / C++)
├── Samples/              # Code samples in C, C++, C#, Java, and Python
├── ADLXPybind/           # Python binding (.pyd) built with pybind11
├── MCPs/                 # MCP servers and client for AI agent integration
└── SDKDoc/               # SDK reference documentation

SDK

The SDK/ directory contains:

• SDK/Include/ — ADLX interface headers (C and C++)
• SDK/ADLXHelper/Windows/C/ — C helper for loading and initializing ADLX
• SDK/ADLXHelper/Windows/Cpp/ — C++ helper for loading and initializing ADLX
• SDK/Platform/ — Windows platform abstraction layer

Samples

Code samples are organized by feature area. Each sample is available in multiple languages:

Category	Description
3DGraphics	Anti-aliasing, tessellation, FSR, AFMF, Radeon Boost, and more
Display	Resolution, color depth, refresh rate, FreeSync, custom resolutions
Desktop	Eyefinity multi-display configuration
GPUTuning	Auto/preset/manual GPU tuning
PerformanceMonitoring	FPS, GPU/CPU metrics, system info
PowerTuning	Power limit, TDP, fan control
MultiMedia	Video super resolution (VSR)
ReceivingEventsNotifications	Event and change notification callbacks
Generic	GPU enumeration, logging, stress test, service calls
I2C	I2C communication
VariableGraphicsMemory	Variable graphics memory (Smart Access Memory)

ADLXPybind

Python bindings for ADLX, exposing the full ADLX interface to Python via pybind11. Produces adlx.pyd and an installable wheel (adlx-*.whl).

MCPs

MCP (Model Context Protocol) servers that expose AMD GPU controls as tools for AI agents and LLM applications.

Component	Description
Server/amd-3dsettings-mcp-server	3D graphics settings (RSR, AFMF, Anti-Lag, Chill, etc.)
Server/amd-display-mcp-server	Display settings (VSR, color, custom resolution)
Server/amd-gputunings-mcp-server	GPU tuning (auto/preset/manual, fan, power, VRAM)
Server/amd-perfmonitoring-mcp-server	GPU/system performance metrics
Client/mcp-integration-with-LLM-router	CLI and Streamlit chatbot client with LLM router

---

Requirements

C / C++ Samples

Item	Version
OS	Windows 10 (32/64-bit) or later
IDE	Visual Studio 2019 with C/C++ components
Build system	CMake 3.8+ (open via VS2019: File > Open > CMake)
AMD driver	ADLX-compatible AMD GPU driver

ADLXPybind (Python Binding)

Item	Version
OS	Windows 11 (32/64-bit)
IDE	Visual Studio 2022 with C/C++ components
Python	3.13
Library	pybind11
Environment variable	PYHOME pointing to Python 3.12 installation

MCPs

Item	Version
OS	Windows 10 (64-bit) or later
Python	3.13
Package manager	uv
AMD driver	ADLX-compatible AMD GPU driver
Python wheel	adlx-1.0.0-py313-none-win_amd64.whl (from ADLXPybind build)

---

Quick Start

1. C++ Sample

# Open in Visual Studio 2019 via File > Open > CMake
# Select Samples/CPP/CMakeLists.txt
# Build configuration: vs2019_Release_x64

2. Python via ADLXPybind

# Build the wheel (requires VS 2022 + Python 3.12 + pybind11)
cd ADLXPybind
# Open ADLXPybind.sln in Visual Studio 2022 and Rebuild
# Or build the wheel:
build_wheel.bat
pip install dist/adlx-*.whl

import adlx
system, _ = adlx.helper.QueryFullVersion()
print(system)

3. MCP Servers (AI Agent Integration)

cd MCPs

# Start a server (example: 3D settings)
cd Server/amd-3dsettings-mcp-server
uv run server.py

# Or start the Streamlit chatbot client
cd Client/mcp-integration-with-LLM-router/adlx_mcp_chatbot
uv run streamlit run app.py

---

License

ADLX SDK is distributed under the license included in ADLX SDK License Agreement.pdf at the root of this repository.

Attribution

• AMD, the AMD Arrow logo, Radeon, Crossfire, and combinations thereof are either registered trademarks or trademarks of Advanced Micro Devices, Inc. in the United States and/or other countries.
• Microsoft, DirectX, Visual Studio, and Windows are either registered trademarks or trademarks of Microsoft Corporation in the United States and/or other countries.