diff --git a/README.md b/README.md
index 99e053bfb..47f2df219 100644
--- a/README.md
+++ b/README.md
@@ -8,7 +8,7 @@
 WanGP supports the Wan (and derived models) but also Hunyuan Video, Flux, Qwen, Z-Image, LongCat, Kandinsky, LTXV, LTX-2, Qwen3 TTS, Chatterbox, HearMula, ... with:
 - Low VRAM requirements (as low as 6 GB of VRAM is sufficient for certain models)
 - Support for old Nvidia GPUs (RTX 10XX, 20xx, ...)
-- Support for AMD GPUs Radeon RX 76XX, 77XX, 78XX & 79XX, instructions in the Installation Section Below.
+- Support for AMD GPUs (RDNA 4, 3, 3.5, and 2), instructions in the Installation Section Below.
 - Very Fast on the latest GPUs
 - Easy to use Full Web based interface
 - Support for many checkpoint Quantized formats: int8, fp8, gguf, NV FP4, Nunchaku
@@ -318,7 +318,7 @@ For detailed installation instructions for different GPU generations:
 
 ### AMD
 For detailed installation instructions for different GPU generations:
-- **[Installation Guide](docs/AMD-INSTALLATION.md)** - Complete setup instructions for Radeon RX 76XX, 77XX, 78XX & 79XX
+- **[Installation Guide](docs/AMD-INSTALLATION.md)** - Complete setup instructions for RDNA 4, 3, 3.5, and 2
 
 ## 🎯 Usage
 
diff --git a/docs/AMD-INSTALLATION.md b/docs/AMD-INSTALLATION.md
index 4f05589eb..8663cb0db 100644
--- a/docs/AMD-INSTALLATION.md
+++ b/docs/AMD-INSTALLATION.md
@@ -1,72 +1,79 @@
-# Installation Guide
+# AMD Installation Guide for Windows (TheRock)
 
-This guide covers installation for specific RDNA3 and RDNA3.5 AMD CPUs (APUs) and GPUs
-running under Windows. 
+This guide covers installation for AMD GPUs and APUs running under Windows using TheRock's official PyTorch wheels.
 
-tl;dr: Radeon RX 7900 GOOD, RX 9700 BAD, RX 6800 BAD. (I know, life isn't fair).
+## Supported GPUs
 
-Currently supported (but not necessary tested):
+Based on [TheRock's official support matrix](https://github.com/ROCm/TheRock/blob/main/SUPPORTED_GPUS.md), the following GPUs are supported on Windows:
 
-**gfx110x**:
+### **gfx110X-all** (RDNA 3):
+* AMD RX 7900 XTX (gfx1100)
+* AMD RX 7800 XT (gfx1101)
+* AMD RX 7700 XT (gfx1101)
+* AMD RX 7700S / Framework Laptop 16 (gfx1102)
+* AMD Radeon 780M Laptop iGPU (gfx1103)
 
-* Radeon RX 7600
-* Radeon RX 7700 XT
-* Radeon RX 7800 XT
-* Radeon RX 7900 GRE
-* Radeon RX 7900 XT
-* Radeon RX 7900 XTX
+### **gfx120X-all** (RDNA 4):
+* AMD RX 9060 XT (gfx1200)
+* AMD RX 9060 (gfx1200)
+* AMD RX 9070 XT (gfx1201)
+* AMD RX 9070 (gfx1201)
 
-**gfx1151**:
+### **gfx1151** (RDNA 3.5 APU):
+* AMD Strix Halo APUs
 
-* Ryzen 7000 series APUs (Phoenix)
-* Ryzen Z1 (e.g., handheld devices like the ROG Ally)
+### **gfx1150** (RDNA 3.5 APU): 
+* AMD Radeon 890M (Ryzen AI 9 HX 370 - Strix Point)
 
-**gfx1201**:
+### Also supported:
+### **gfx103X-dgpu**: (RDNA 2)
 
-* Ryzen 8000 series APUs (Strix Point) 
-* A [frame.work](https://frame.work/au/en/desktop) desktop/laptop
+<br>
 
+> **Note:** If your GPU is not listed above, it may not be supported by TheRock on Windows. Support status and future updates can be found in the [official documentation](https://github.com/ROCm/TheRock/blob/main/SUPPORTED_GPUS.md).
 
 ## Requirements
 
-- Python 3.11 (3.12 might work, 3.10 definately will not!)
+- Python 3.11 (recommended for Wan2GP - TheRock currently supports Python 3.11, 3.12, and 3.13).
+- Windows 10/11
 
 ## Installation Environment
 
-This installation uses PyTorch 2.7.0 because that's what currently available in
-terms of pre-compiled wheels.
+This installation uses PyTorch wheels built by TheRock.
 
 ### Installing Python
 
-Download Python 3.11 from [python.org/downloads/windows](https://www.python.org/downloads/windows/). Hit Ctrl+F and search for "3.11". Dont use this direct link: [https://www.python.org/ftp/python/3.11.9/python-3.11.9-amd64.exe](https://www.python.org/ftp/python/3.11.9/python-3.11.9-amd64.exe) -- that was an IQ test.
+Download Python 3.11 from [python.org/downloads/windows](https://www.python.org/downloads/windows/). Press Ctrl+F and search for "3.11." to find the newest version available for installation.
 
-After installing, make sure `python --version` works in your terminal and returns 3.11.x
+Alternatively, you can use this direct link: [Python 3.11.9 (64-bit)](https://www.python.org/ftp/python/3.11.9/python-3.11.9-amd64.exe).
 
-If not, you probably need to fix your PATH. Go to:
+After installing, make sure `python --version` works in your terminal and returns `3.11.9`
 
-* Windows + Pause/Break
-* Advanced System Settings
-* Environment Variables
-* Edit your `Path` under User Variables
+If it doesn’t, you need to add Python to your PATH:
 
-Example correct entries:
+* Press the `Windows` key, type `Environment Variables`, and select `Edit the system environment variables`.
+* In the `System Properties` window, click `Environment Variables…`.
+* Under `User variables`, find `Path`, then click `Edit` → `New` and add the following entries (replace `<username>` with your Windows username):
 
 ```cmd
-C:\Users\YOURNAME\AppData\Local\Programs\Python\Launcher\
-C:\Users\YOURNAME\AppData\Local\Programs\Python\Python311\Scripts\
-C:\Users\YOURNAME\AppData\Local\Programs\Python\Python311\
+C:\Users\<username>\AppData\Local\Programs\Python\Launcher\
+C:\Users\<username>\AppData\Local\Programs\Python\Python311\Scripts\
+C:\Users\<username>\AppData\Local\Programs\Python\Python311\
 ```
 
-If that doesnt work, scream into a bucket.
+> **Note:** If Python still doesn't show the correct version after updating PATH, try signing out and signing back in to Windows to apply the changes.
 
 ### Installing Git
 
-Get Git from [git-scm.com/downloads/win](https://git-scm.com/downloads/win). Default install is fine.
+Download Git from [git-scm.com/downloads/windows](https://git-scm.com/install/windows) and install it. The default installation options are fine.
 
 
-## Install (Windows, using `venv`)
+## Installation Steps (Windows, using a Python `venv`)
+> **Note:** The following commands are intended for use in the Windows Command Prompt (CMD).  
+> If you are using PowerShell, some commands (like comments and activating the virtual environment) may differ.
 
-### Step 1: Download and Set Up Environment
+
+### Step 1: Download and set up Wan2GP Environment
 
 ```cmd
 :: Navigate to your desired install directory
@@ -76,71 +83,192 @@ cd \your-path-to-wan2gp
 git clone https://github.com/deepbeepmeep/Wan2GP.git
 cd Wan2GP
 
-:: Create virtual environment using Python 3.10.9
+:: Create virtual environment
 python -m venv wan2gp-env
 
 :: Activate the virtual environment
 wan2gp-env\Scripts\activate
 ```
 
-### Step 2: Install PyTorch
+> **Note:** If you have multiple versions of Python installed, use `py -3.11 -m venv wan2gp-env` instead of `python -m venv wan2gp-env` to ensure the correct version is used.
+
+### Step 2: Install ROCm/PyTorch by TheRock
+
+**IMPORTANT:** Choose the correct index URL for your GPU family!
+
+#### For gfx110X-all (RX 7900 XTX, RX 7800 XT, etc.):
 
-The pre-compiled wheels you need are hosted at [scottt's rocm-TheRock releases](https://github.com/scottt/rocm-TheRock/releases). Find the heading that says:
+```cmd
+pip install --pre torch torchaudio torchvision rocm[devel] --index-url https://rocm.nightlies.amd.com/v2/gfx110X-all/
+```
 
-**Pytorch wheels for gfx110x, gfx1151, and gfx1201**
+#### For gfx120X-all (RX 9060, RX 9070, etc.):
 
-Don't click this link: [https://github.com/scottt/rocm-TheRock/releases/tag/v6.5.0rc-pytorch-gfx110x](https://github.com/scottt/rocm-TheRock/releases/tag/v6.5.0rc-pytorch-gfx110x). It's just here to check if you're skimming.
+```cmd
+pip install --pre torch torchaudio torchvision rocm[devel] --index-url https://rocm.nightlies.amd.com/v2/gfx120X-all/
+```
+
+#### For gfx1151 (Strix Halo iGPU):
+
+```cmd
+pip install --pre torch torchaudio torchvision rocm[devel] --index-url https://rocm.nightlies.amd.com/v2/gfx1151/
+```
+
+#### For gfx1150 (Radeon 890M - Strix Point):
+
+```cmd
+pip install --pre torch torchaudio torchvision rocm[devel] --index-url https://rocm.nightlies.amd.com/v2-staging/gfx1150/
+```
 
-Copy the links of the closest binaries to the ones in the example below (adjust if you're not running Python 3.11), then hit enter.
+#### For gfx103X-dgpu (RDNA 2):
 
 ```cmd
-pip install ^
-    https://github.com/scottt/rocm-TheRock/releases/download/v6.5.0rc-pytorch-gfx110x/torch-2.7.0a0+rocm_git3f903c3-cp311-cp311-win_amd64.whl ^
-    https://github.com/scottt/rocm-TheRock/releases/download/v6.5.0rc-pytorch-gfx110x/torchaudio-2.7.0a0+52638ef-cp311-cp311-win_amd64.whl ^
-    https://github.com/scottt/rocm-TheRock/releases/download/v6.5.0rc-pytorch-gfx110x/torchvision-0.22.0+9eb57cd-cp311-cp311-win_amd64.whl
+pip install --pre torch torchaudio torchvision rocm[devel] --index-url https://rocm.nightlies.amd.com/v2-staging/gfx103X-dgpu/
 ```
 
-### Step 3: Install Dependencies
+This will automatically install the latest PyTorch, torchaudio, and torchvision wheels with ROCm support.
+
+### Step 3: Install Wan2GP Dependencies
 
 ```cmd
 :: Install core dependencies
 pip install -r requirements.txt
 ```
 
+### Step 4: Verify Installation
+
+```cmd
+python -c "import torch; print('PyTorch:', torch.__version__); print('ROCm available:', torch.cuda.is_available()); print('Device:', torch.cuda.get_device_name(0) if torch.cuda.is_available() else 'No GPU')"
+```
+
+Expected output example:
+```
+PyTorch: 2.11.0+rocm7.12.0
+ROCm available: True
+Device: AMD Radeon RX 9070 XT
+```
+
 ## Attention Modes
 
-WanGP supports several attention implementations, only one of which will work for you:
+WanGP supports multiple attention implementations via [triton-windows](https://github.com/woct0rdho/triton-windows/).
 
-- **SDPA** (default): Available by default with PyTorch.  This uses the built-in aotriton accel library, so is actually pretty fast.
+First, install `triton-windows` in your virtual environment.
+If you have an older version of Triton installed, uninstall it first.
+ROCm SDK needs to be initialized.
+Visual Studio environment should also be activated.
 
-## Performance Profiles
+```cmd
+pip uninstall triton
+pip install triton-windows
+rocm-sdk init
+"C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Auxiliary\Build\vcvars64.bat" >nul 2>&1
+```
 
-Choose a profile based on your hardware:
+### Supported attention implementations
 
-- **Profile 3 (LowRAM_HighVRAM)**: Loads entire model in VRAM, requires 24GB VRAM for 8-bit quantized 14B model
-- **Profile 4 (LowRAM_LowVRAM)**: Default, loads model parts as needed, slower but lower VRAM requirement
+- **SageAttention V1** (Requires the `.post26` wheel or newer to fix Triton compilation issues without needing unofficial patches. Download it from [this](https://github.com/Comfy-Org/wheels/actions/runs/21343435018) URL)
+
+```cmd
+pip install "sageattention <2"
+```
+
+- **FlashAttention-2** (Only the Triton backend is supported): 
+```cmd
+git clone https://github.com/Dao-AILab/flash-attention.git
+cd flash-attention
+pip install ninja
+pip install packaging
+set FLASH_ATTENTION_TRITON_AMD_ENABLE=TRUE && python setup.py install
+```
+
+- **SDPA Flash**: Available by default in PyTorch on post-RDNA2 GPUs via AOTriton.
 
 ## Running Wan2GP
 
-In future, you will have to do this:
+For future sessions, activate the environment every time if it isn't already activated, then run `python wgp.py`:
 
 ```cmd
-cd \path-to\wan2gp
-wan2gp\Scripts\activate.bat
+cd \path-to\Wan2GP
+wan2gp-env\Scripts\activate
+:: Add the AMD-specific environment variables mentioned below here
 python wgp.py
 ```
 
-For now, you should just be able to type `python wgp.py` (because you're already in the virtual environment)
+It is advised to set the following environment variables at the start of every new session (you can create a `.bat` file that activates your venv, sets these, then launches `wgp.py`):
+
+```cmd
+set ROCM_HOME=%ROCM_ROOT%
+set PATH=%ROCM_ROOT%\lib\llvm\bin;%ROCM_BIN%;%PATH%
+set CC=clang-cl
+set CXX=clang-cl
+set DISTUTILS_USE_SDK=1
+set FLASH_ATTENTION_TRITON_AMD_ENABLE=TRUE
+set TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1
+```
+
+MIOpen (AMD’s equivalent of NVIDIA’s cuDNN) is not yet fully stable on several architectures; it can cause out-of-memory errors (OOMs), crash the display driver, or significantly increase generation times. Currently, it is recommended to either use fast mode by setting:
+
+```cmd
+set MIOPEN_FIND_MODE=FAST
+```
+
+Alternatively, you can disable MIOpen entirely by editing `wgp.py` and adding the following line below `import torch` (around line 51):
+
+```cmd
+...
+:: /Lines already in the file/
+:: import torch
+torch.backends.cudnn.enabled = False # <-- Add this here
+:: import gc
+...
+```
+
+To verify that it is disabled, or to enable verbose logging, you can set:
+
+```cmd
+set MIOPEN_ENABLE_LOGGING=1
+set MIOPEN_ENABLE_LOGGING_CMD=1
+set MIOPEN_LOG_LEVEL=5
+```
 
 ## Troubleshooting
 
-- If you use a HIGH VRAM mode, don't be a fool.  Make sure you use VAE Tiled Decoding.
+### GPU Not Detected
+
+If `torch.cuda.is_available()` returns `False`:
+
+1. **Verify your GPU is supported** - Check the [Supported GPUs](#supported-gpus) list above
+2. **Check AMD drivers** - Ensure you have the latest AMD Adrenalin drivers installed
+3. **Verify correct index URL** - Make sure you used the right GPU family index URL
+
+### Installation Errors
+
+**"Could not find a version that satisfies the requirement":**
+- Double-check that you're using the correct `--index-url` for your GPU family. You can also try adding the `--pre` flag or replacing `/v2/` in the URL with `/v2/staging/`
+- Ensure you're using Python 3.11, and not 3.10
+- Try adding `--pre` flag if not already present
+
+**"No matching distribution found":**
+- Your GPU architecture may not be supported
+- Check that you've activated your virtual environment
+
+### Performance Issues
+
+- **Monitor VRAM usage** - Reduce batch size or resolution if running out of memory
+- **Close GPU-intensive apps** - Apps with hardware acceleration enabled (browsers, Discord etc.).
+
+### Known Issues
+
+Windows packages are new and may be unstable.
+
+Known issues are tracked at: https://github.com/ROCm/TheRock/issues/808
 
-### Memory Issues
+## Additional Resources
 
-- Use lower resolution or shorter videos
-- Enable quantization (default)
-- Use Profile 4 for lower VRAM usage
-- Consider using 1.3B models instead of 14B models
+- [TheRock GitHub Repository](https://github.com/ROCm/TheRock/)
+- [Releases Documentation](https://github.com/ROCm/TheRock/blob/main/RELEASES.md)
+- [Supported GPU Architectures](https://github.com/ROCm/TheRock/blob/main/SUPPORTED_GPUS.md)
+- [Roadmap](https://github.com/ROCm/TheRock/blob/main/ROADMAP.md)
+- [ROCm Documentation](https://rocm.docs.amd.com/)
 
-For more troubleshooting, see [TROUBLESHOOTING.md](TROUBLESHOOTING.md) 
+For additional troubleshooting guidance for Wan2GP, see [TROUBLESHOOTING.md](https://github.com/deepbeepmeep/Wan2GP/blob/main/docs/TROUBLESHOOTING.md).