Install AMD ROCm 7.9.0#
2025-10-20
14 min read time
Use the following selector to choose your installation method for your supported AMD GPU or APU and operating system. For information about supported operating systems and compatible AMD devices, see the Compatibility matrix.
Prerequisites#
Install Python 3.11.
sudo apt install python3.11 python3.11-venv
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Install Python 3.12 or 3.13.
sudo apt install python3.12 python3.12-venv
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Register your Enterprise Linux.
subscription-manager register --username <username> --password <password> sudo subscription-manager attach --auto
Update your Enterprise Linux.
sudo dnf update --releasever=9.6 --exclude=\*release\*
Install Python 3.11, 3.12 or 3.13.
sudo dnf install python3.11 python3.11-pip
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Register your Enterprise Linux.
subscription-manager register --username <username> --password <password>
Update your Enterprise Linux.
sudo dnf update --releasever=10.0 --exclude=\*release\*
Install Python 3.12 or 3.13.
sudo dnf install python3.12 python3.12-pip
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Install Python 3.12 or 3.13.
sudo apt install python3.12 python3.12-venv
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Remove any existing HIP SDK installations and other conflicting AMD graphics software.
Install the Adrenalin Driver version 25.9.2. For instructions, see Install AMD Software: Adrenalin Edition.
Install a supported Python version: 3.11, 3.12, or 3.13.
Prerequisites#
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Register your Enterprise Linux.
subscription-manager register --username <username> --password <password> sudo subscription-manager attach --auto
Update your Enterprise Linux.
sudo dnf update --releasever=9.6 --exclude=\*release\*
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Register your Enterprise Linux.
subscription-manager register --username <username> --password <password>
Update your Enterprise Linux.
sudo dnf update --releasever=10.0 --exclude=\*release\*
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Configure permissions for GPU access.
sudo usermod -a -G render,video $LOGNAME # Add the current user to the render and video groups
Note
To apply all settings, reboot your system.
Remove any existing HIP SDK installations and other conflicting AMD graphics software.
Install the Adrenalin Driver version 25.9.2. For instructions, see Install AMD Software: Adrenalin Edition.
Installing#
Install AMD GPU Driver#
For information about driver compatibility, see the ROCm compatibility matrix.
For information about the AMD GPU driver installation, see the Ubuntu native installation in the AMD Instinct Data Center GPU Documentation.
Install AMD GPU Driver#
For information about driver compatibility, see the ROCm compatibility matrix.
For information about the AMD GPU driver installation, see the RHEL native installation in the AMD Instinct Data Center GPU Documentation.
Installing#
Install kernel driver#
Supported Ryzen AI APUs require the inbox kernel driver included with Ubuntu 24.04.3.
Install ROCm#
Set up your Python virtual environment.
python3.11 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.11 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx950-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.11 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.11 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx94X-dcgpu/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv source .venv/bin/activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ "rocm[libraries,devel]"
Set up your Python virtual environment.
python3.12 -m venv .venv .venv\Scripts\activate
Install ROCm wheels packages.
python -m pip install --index-url https://repo.amd.com/rocm/whl/gfx1151/ "rocm[libraries,devel]"
Install ROCm#
Create the installation directory. For example:
mkdir therock-tarball && cd therock-tarball
Note
Subsequent commands assume you’re working with the
therock-tarball
directory. If you choose a different directory name, adjust the subsequent commands accordingly.Download and unpack the tarball.
wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx950-dcgpu-7.9.0rc1.tar.gz mkdir install tar -xf *.tar.gz -C install
Create the installation directory. For example:
mkdir therock-tarball && cd therock-tarball
Note
Subsequent commands assume you’re working with the
therock-tarball
directory. If you choose a different directory name, adjust the subsequent commands accordingly.Download and unpack the tarball.
wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx94X-dcgpu-7.9.0rc1.tar.gz mkdir install tar -xf *.tar.gz -C install
Create the installation directory. For example:
mkdir therock-tarball && cd therock-tarball
Note
Subsequent commands assume you’re working with the
therock-tarball
directory. If you choose a different directory name, adjust the subsequent commands accordingly.Download and unpack the tarball.
wget https://repo.amd.com/rocm/tarball/therock-dist-linux-gfx1151-7.9.0rc1.tar.gz mkdir install tar -xf *.tar.gz -C install
Important
Do not copy/replace the ROCm-SDK compiler and runtime DLLs to
System32
as this can cause conflicts.Disable the following Windows security features as they can interfere with ROCm functionality:
Turn off WDAG (Windows Defender Application Guard)
Control Panel > Programs > Programs and Features > Turn Windows features on or off > Deselect “Microsoft Defender Application Guard”
Turn off SAC (Smart App Control)
Settings > Privacy & security > Windows Security > App & browser control > Smart App Control settings > Off
Create the installation directory in
C:\TheRock\build
.Note
Subsequent commands assume you’re working with the
C:\TheRock\build
directory.Download the tarball and extract the contents to
C:\TheRock\build
.Set the following environment variables using the command prompt as an administrator:
setx HIP_DEVICE_LIB_PATH “C:\TheRock\build\lib\llvm\amdgcn\bitcode” /M setx HIP_PATH “C:\TheRock\build” /M setx HIP_PLATFORM “amd” /M setx LLVM_PATH “C:\TheRock\build\lib\llvm” /M
Add the following paths into PATH environment variable using your system settings GUI.
C:\TheRock\build\bin
C:\TheRock\build\lib\llvm\bin
Open a new command prompt window for the environment variables to take effect. Run
set
to see the list of active variables.set
Post-installation#
Configure ROCm PATH. Make sure you’re in the
therock-tarball
directory before proceeding.export ROCM_PATH=$PWD export PATH=$PATH:$ROCM_PATH/install/bin
Configure
LD_LIBRARY_PATH
.export LD_LIBRARY_PATH=$ROCM_PATH/install/lib
Verify the ROCm installation.
rocminfo amd-smi
Example output of
rocminfo
ROCk module is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.18 Runtime Ext Version: 1.14 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED XNACK enabled: NO DMAbuf Support: YES VMM Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Uuid: CPU-XX Marketing Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Vendor Name: CPU [output truncated]
Configure ROCm PATH. Make sure you’re in the
therock-tarball
directory before proceeding.export ROCM_PATH=$PWD export PATH=$PATH:$ROCM_PATH/install/bin
Configure
LD_LIBRARY_PATH
.export LD_LIBRARY_PATH=$ROCM_PATH/install/lib
Verify the ROCm installation.
rocminfo amd-smi
Example output of
rocminfo
ROCk module is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.18 Runtime Ext Version: 1.14 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED XNACK enabled: NO DMAbuf Support: YES VMM Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Uuid: CPU-XX Marketing Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Vendor Name: CPU [output truncated]
Verify the ROCm installation.
hipinfo
Example output of hipinfo
--------------------------------------------------------------------------------
device# 0
Name: AMD Radeon(TM) 8060S Graphics
pciBusID: 197
pciDeviceID: 0
pciDomainID: 0
multiProcessorCount: 20
[output truncated]
Post-installation#
Verify the ROCm installation.
rocminfo amd-smi
Example output of
rocminfo
ROCk module is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.18 Runtime Ext Version: 1.14 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED XNACK enabled: NO DMAbuf Support: YES VMM Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Uuid: CPU-XX Marketing Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Vendor Name: CPU [output truncated]
Inspect your ROCm installation in your Python environment.
pip freeze | grep rocm which rocm-sdk ls .venv/bin
Test your ROCm installation.
rocm-sdk targets rocm-sdk path --cmake rocm-sdk path --bin rocm-sdk path --root rocm-sdk test
To learn more about the
rocm-sdk
tool and to see example expected outputs, see Using ROCm Python packages (TheRock).
Tip
If you need to deactivate your Python virtual environment when finished, run:
deactivate
Verify the ROCm installation.
rocminfo amd-smi
Example output of
rocminfo
ROCk module is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.18 Runtime Ext Version: 1.14 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED XNACK enabled: NO DMAbuf Support: YES VMM Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Uuid: CPU-XX Marketing Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Vendor Name: CPU [output truncated]
Inspect your ROCm installation in your Python environment.
pip freeze | grep rocm which rocm-sdk ls .venv/bin
Test your ROCm installation.
rocm-sdk targets rocm-sdk path --cmake rocm-sdk path --bin rocm-sdk path --root rocm-sdk test
To learn more about the
rocm-sdk
tool and to see example expected outputs, see Using ROCm Python packages (TheRock).
Tip
If you need to deactivate your Python virtual environment when finished, run:
deactivate
Verify the ROCm installation.
rocminfo amd-smi
Example output of
rocminfo
ROCk module is loaded ===================== HSA System Attributes ===================== Runtime Version: 1.18 Runtime Ext Version: 1.14 System Timestamp Freq.: 1000.000000MHz Sig. Max Wait Duration: 18446744073709551615 (0xFFFFFFFFFFFFFFFF) (timestamp count) Machine Model: LARGE System Endianness: LITTLE Mwaitx: DISABLED XNACK enabled: NO DMAbuf Support: YES VMM Support: YES ========== HSA Agents ========== ******* Agent 1 ******* Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Uuid: CPU-XX Marketing Name: AMD RYZEN AI MAX+ PRO 395 w/ Radeon 8060S Vendor Name: CPU [output truncated]
Inspect your ROCm installation in your Python environment.
pip freeze | grep rocm which rocm-sdk ls .venv/bin
Test your ROCm installation.
rocm-sdk targets rocm-sdk path --cmake rocm-sdk path --bin rocm-sdk path --root rocm-sdk test
To learn more about the
rocm-sdk
tool and to see example expected outputs, see Using ROCm Python packages (TheRock).
Tip
If you need to deactivate your Python virtual environment when finished, run:
deactivate
Verify the ROCm installation.
hipinfo
Example output of
hipinfo
-------------------------------------------------------------------------------- device# 0 Name: AMD Radeon(TM) 8060S Graphics pciBusID: 197 pciDeviceID: 0 pciDomainID: 0 multiProcessorCount: 20 [output truncated]
Inspect your ROCm installation in your Python environment.
pip freeze where rocm-sdk dir .venv\Scripts
Test your ROCm installation.
rocm-sdk test
To learn more about the
rocm-sdk
tool and to see example expected outputs, see Using ROCm Python packages (TheRock).
Tip
If you need to deactivate your Python virtual environment when finished, run:
deactivate
Uninstalling#
To uninstall ROCm, remove your installation directory.
sudo rm -rf therock-tarball
To uninstall ROCm, remove your installation directory.
sudo rm -rf therock-tarball
Delete the
C:\TheRock\build
and its contents.Delete the environment variables. For example, using Powershell as an administrator:
[Environment]::SetEnvironmentVariable("HIP_PATH", $null, "Machine")
[Environment]::SetEnvironmentVariable("HIP_DEVICE_LIB_PATH", $null, "Machine")
[Environment]::SetEnvironmentVariable("HIP_PLATFORM", $null, "Machine")
[Environment]::SetEnvironmentVariable("LLVM_PATH", $null, "Machine")
Remove the following paths from your PATH environment variable using your system settings GUI.
C:\TheRock\build\bin
C:\TheRock\build\lib\llvm\bin
If you want to uninstall the Adrenalin driver, see Uninstall AMD Software.
Uninstalling#
Clear the pip cache.
sudo rm -rf ~/.cache/pip
Remove your local Python virtual environment.
sudo rm -rf .venv
Clear the pip cache.
sudo rm -rf ~/.cache/pip
Remove your local Python virtual environment.
sudo rm -rf .venv
Clear the pip cache.
pip cache purge
Remove your local Python virtual environment.
rmdir /s /q .venv
Troubleshooting#
Additional packages for Docker installation#
Docker images often only include a minimal set of installations, meaning some essential packages might be missing. When installing ROCm within a Docker container, you might need to install additional packages for a successful installation. Use the following commands to install the prerequisite packages.
apt update
apt install sudo wget
Additional packages for Docker installation#
Docker images often only include a minimal set of installations, meaning some essential packages might be missing. When installing ROCm within a Docker container, you might need to install additional packages for a successful installation. Use the following commands to install the prerequisite packages.
dnf install sudo libatomic
Troubleshooting#
Additional packages for Docker installation#
Docker images often only include a minimal set of installations, meaning some essential packages might be missing. When installing ROCm within a Docker container, you might need to install additional packages for a successful installation. Use the following commands to install the prerequisite packages.
apt update
apt install sudo wget