ROCm Installation¶

Install AMD ROCm stack natively on Ubuntu 24.04 for the AMD Ryzen AI Max+ 395 APU.

APU Support Status¶

Strix Point Support

As of early 2025, ROCm support for Strix Point APUs (gfx1151) is evolving. Check AMD ROCm documentation for the latest compatibility matrix.

Current Compatibility¶

Component	Support Level	Notes
amdgpu kernel driver	Good	Included in Ubuntu 24.04 kernel
ROCm runtime	Experimental	May require `HSA_OVERRIDE_GFX_VERSION`
HIP	Experimental	Some applications work
OpenCL	Good	Generally functional

The AMD Ryzen AI Max+ 395 uses the RDNA 3.5 architecture with GPU ID gfx1151. Official ROCm support typically lags new hardware releases.

APU vs Discrete GPU¶

Aspect	APU	Discrete GPU
ROCm support	Newer, less tested	Mature
Memory	Shared system RAM	Dedicated VRAM
Device nodes	`/dev/kfd`, `/dev/dri`	Same
Performance	Memory bandwidth limited	Higher bandwidth

Prerequisites¶

Kernel Requirements¶

Ubuntu 24.04 includes a sufficiently recent kernel. Verify:

uname -r
# Should be 6.8 or newer

Check GPU Detection¶

Verify the APU is recognized:

# Check for AMD GPU
lspci | grep -i vga
# Output should include AMD Radeon

# Check DRI devices
ls -la /dev/dri/
# Should show card0 and renderD128

Required Groups¶

# Add user to required groups
sudo usermod -aG video,render $USER

# Apply group changes (or log out and back in)
newgrp video
newgrp render

Installation Methods¶

Method 1: amdgpu-install (Recommended)¶

The amdgpu-install script provides the simplest installation path.

Download the installer:

# Ubuntu 24.04 (Noble)
wget https://repo.radeon.com/amdgpu-install/latest/ubuntu/noble/amdgpu-install_6.3.60300-1_all.deb

# Install the installer package
sudo apt install ./amdgpu-install_6.3.60300-1_all.deb

Version Numbers

The version (6.3.60300-1) changes with ROCm releases. Check repo.radeon.com for the latest.

Install ROCm:

# Install ROCm with all common components
sudo amdgpu-install --usecase=rocm

# Or for minimal installation
sudo amdgpu-install --usecase=rocm --no-dkms

Available use cases:

Use Case	Components
`rocm`	Full ROCm stack
`graphics`	Graphics drivers only
`opencl`	OpenCL runtime
`hip`	HIP development
`rocmdev`	ROCm development tools

Method 2: Manual Installation¶

For more control over components:

Add AMD repository:

# Import GPG key
wget -qO - https://repo.radeon.com/rocm/rocm.gpg.key | sudo gpg --dearmor -o /etc/apt/keyrings/rocm.gpg

# Add repository
echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/6.3 noble main" | sudo tee /etc/apt/sources.list.d/rocm.list

# Update package lists
sudo apt update

Install components:

# Core ROCm runtime
sudo apt install rocm-hip-runtime

# ROCm SMI for monitoring
sudo apt install rocm-smi-lib

# Development tools (optional)
sudo apt install rocm-dev

# OpenCL runtime
sudo apt install rocm-opencl-runtime

Environment Configuration¶

Path Setup¶

Add ROCm to your PATH:

# Add to ~/.bashrc or ~/.profile
echo 'export PATH=$PATH:/opt/rocm/bin' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/rocm/lib' >> ~/.bashrc

# Apply changes
source ~/.bashrc

APU-Specific Variables¶

For Strix Point APUs, you may need to override the GPU version:

# If ROCm doesn't recognize gfx1151, try gfx1100 (RDNA 3)
export HSA_OVERRIDE_GFX_VERSION=11.0.0

# Add to ~/.bashrc for persistence
echo 'export HSA_OVERRIDE_GFX_VERSION=11.0.0' >> ~/.bashrc

Version Override

The HSA_OVERRIDE_GFX_VERSION trick tells ROCm to treat your GPU as a different (supported) architecture. This may cause instability but often enables functionality on newer hardware.

Other Useful Variables¶

Variable	Purpose	Example
`HSA_OVERRIDE_GFX_VERSION`	Force GPU architecture	`11.0.0`
`HIP_VISIBLE_DEVICES`	Limit visible GPUs	`0`
`ROCR_VISIBLE_DEVICES`	Alternative device selection	`0`
`GPU_MAX_HW_QUEUES`	Hardware queue limit	`8`

Verification¶

rocminfo¶

Check ROCm detects the GPU:

rocminfo

Expected output includes:

ROCk module is loaded
HSA System Attributes:
  ...
Agent 1:
  Name:                    gfx1151
  Marketing Name:          AMD Radeon Graphics
  ...
  Pool Info:
    Segment:               GLOBAL; FLAGS: FINE GRAINED
    Size:                  XX(XXX)KB

If you see gfx1151 or your overridden version, ROCm detected the APU.

rocm-smi¶

Monitor GPU status:

# Basic status
rocm-smi

# Detailed info
rocm-smi --showallinfo

# Memory usage
rocm-smi --showmeminfo vram

# Watch in real-time
watch -n 1 rocm-smi

clinfo (OpenCL)¶

Verify OpenCL functionality:

# Install clinfo if needed
sudo apt install clinfo

# Check OpenCL devices
clinfo

Look for your AMD device in the output.

Simple GPU Test¶

Run a basic HIP test:

# Install ROCm examples
sudo apt install rocm-hip-sdk

# Run device query
/opt/rocm/bin/hipInfo

Troubleshooting¶

GPU Not Detected¶

# Check if amdgpu module is loaded
lsmod | grep amdgpu

# If not loaded, try loading it
sudo modprobe amdgpu

# Check for errors
dmesg | grep -i amdgpu

Permission Denied¶

# Verify group membership
groups $USER
# Should include: video render

# Check device permissions
ls -la /dev/kfd /dev/dri/*

# If permissions wrong, add udev rule
echo 'KERNEL=="kfd", GROUP="render", MODE="0660"' | sudo tee /etc/udev/rules.d/70-kfd.rules
sudo udevadm control --reload-rules
sudo udevadm trigger

HSA Error: Invalid Code Object¶

This typically means architecture mismatch:

# Try different HSA override versions
export HSA_OVERRIDE_GFX_VERSION=11.0.0  # RDNA 3
# or
export HSA_OVERRIDE_GFX_VERSION=11.0.1  # RDNA 3.5 variant

ROCm Version Mismatch¶

# Check installed version
apt list --installed | grep rocm

# Remove conflicting versions
sudo amdgpu-install --uninstall

# Reinstall specific version
sudo amdgpu-install --usecase=rocm

Kernel Module Issues¶

# Check DKMS status
dkms status

# Rebuild modules if needed
sudo dkms autoinstall

# Or reinstall amdgpu-dkms
sudo apt install --reinstall amdgpu-dkms

Integration with AI Frameworks¶

llama.cpp with ROCm¶

Build llama.cpp with HIP support:

git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp

# Build with HIP (ROCm)
make GGML_HIP=1

# For APU, you may need
HSA_OVERRIDE_GFX_VERSION=11.0.0 make GGML_HIP=1

Ollama with ROCm¶

Ollama includes ROCm support:

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Run with ROCm (detected automatically if installed)
ollama run llama3.2

Check Ollama is using GPU:

ollama ps
# Should show GPU memory usage

PyTorch with ROCm¶

Install PyTorch with ROCm support:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm6.0

Verify:

import torch
print(torch.cuda.is_available())  # Uses HIP, still returns True
print(torch.cuda.device_count())
print(torch.cuda.get_device_name(0))

Updating ROCm¶

See Driver Updates for procedures on keeping ROCm current.