iTranslated by AI
The content below is an AI-generated translation. This is an experimental feature, and may contain errors. View original article
🌟
Strix Halo (Ubuntu 25.10) + Distrobox (Ubuntu 24.04) ROCm Setup Guide
Overview
This document is a procedural guide for setting up ROCm and the ComfyUI execution environment on machines equipped with AMD Strix Halo (Ryzen AI MAX+ 395).
Since a recent kernel is required, Ubuntu 25.10 is used as the host, while a Ubuntu 24.04 environment is utilized via Distrobox to ensure the stable operation of ROCm.
I also referred to the following guide:
Basic Policy
- Host OS: Ubuntu 25.10 (to apply the latest Kernel / Firmware)
- Guest OS: Ubuntu 24.04 (built with Distrobox for official ROCm support)
-
Advantages:
- Ease of Environment Setup: The environment can be built with a single command without cluttering the host OS.
- Hardware Integration: Unlike Docker, no GPU passthrough configuration is required.
- Workflow Efficiency: Data handling is easy due to home directory sharing.
1. Checking the Host Environment
Here is the host environment information at the start of the process.
$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 25.10
Release: 25.10
Codename: questing
$ uname -a
Linux mss1 6.18.6-061806-generic #202601171700 SMP PREEMPT_DYNAMIC Sat Jan 17 19:29:30 UTC 2026 x86_64 GNU/Linux
We will configure ComfyUI later, but let's prepare it now.
This can be executed on either the host or the container.
Just keep the location in mind.
I proceeded with the work in ~/ComfyUI for the time being.
# Install Git if not present
$ sudo apt update && sudo apt install git -y
$ git clone https://github.com/Comfy-Org/ComfyUI
2. Creating the Container (Distrobox)
Create a Ubuntu 24.04 container to use as the working environment.
# Create container
$ distrobox create -i ubuntu:24.04 -n strix-ai
# Enter container
$ distrobox enter strix-ai
3. [Host Side] GPU Performance Settings
Note: This step must be performed on the host OS side. Since GUI tools (like CoreCtrl) are environment-dependent, we will configure settings by directly manipulating sysfs.
Procedure
First, check the default status.
# Check performance level (no configuration needed if it is already "high")
$ cat /sys/class/drm/card*/device/power_dpm_force_performance_level
auto
# Real-time check of current GPU clock
$ watch -n 1 "cat /sys/class/drm/card*/device/pp_dpm_sclk"
Every 1.0s: cat /sys/class/drm/card*/device/pp_dpm_sclk mss1: Wed Jan 21 09:36:31 2026
0: 600Mhz *
1: 1100Mhz
2: 2900Mhz
Manually change to Performance Mode and confirm that the clock speed increases.
# First, check which number is assigned to the APU (Strix Halo)
$ ls /sys/class/drm/card*/device/power_dpm_force_performance_level
/sys/class/drm/card0/device/power_dpm_force_performance_level
# For card0 (please adjust according to your environment)
$ echo "high" | sudo tee /sys/class/drm/card0/device/power_dpm_force_performance_level
Here is the confirmation log after the setting.
# Check performance level
$ cat /sys/class/drm/card*/device/power_dpm_force_performance_level
high
# Real-time check of current GPU clock (confirm it stabilizes at 2900MHz)
$ watch -n 1 "cat /sys/class/drm/card*/device/pp_dpm_sclk"
Every 1.0s: cat /sys/class/drm/card*/device/pp_dpm_sclk mss1: Wed Jan 21 09:40:06 2026
0: 2900Mhz
1: 1100Mhz
2: 2900Mhz *
Reference: About BIOS Settings
BIOS settings determine the hardware's power ceiling (TDP).
- Balance Mode: Standard power consumption (around 45W~65W). Prioritizes quietness.
- Performance Mode: Expands the TDP limit (80W~120W+). Recommended for drawing out the full performance of the Strix Halo (especially when handling heavy tasks that utilize the 128GB shared memory).
4. [Container Side] Environment Setup
From here on, the work will be performed inside the Distrobox container (strix-ai).
4.1 Improving the Prompt Display (Recommended)
To make it easier to distinguish between inside and outside the container, add the following to .bashrc.
$ vi ~/.bashrc
# Add to the end of .bashrc
# Execute only when inside Distrobox (container)
if [ -f "/run/.containerenv" ]; then
# 1. Get container name
CONTAINER_NAME=$(grep "name=" /run/.containerenv | cut -d'"' -f2)
# 2. Get ROCm version (repurposing existing logic)
ROCM_VER=$(cat /opt/rocm-*/.info/version 2>/dev/null | head -n 1 | cut -d. -f1,2)
if [ -z "$ROCM_VER" ]; then ROCM_VER="no-rocm"; fi
# 3. Display both in the prompt (e.g., (strix-ai:6.3) hisayan@...)
# Using colors like bold blue for the container name and magenta for ROCm makes it easier to read.
export PS1="\[\e[1;34m\]($CONTAINER_NAME:\[\e[1;35m\]$ROCM_VER\[\e[1;34m\])\[\e[m\] ${PS1}"
fi
4.2 Installing ROCm and the Python Environment
Add the ROCm repository and install the necessary packages. Also, create a Python 3.12 virtual environment.
# Add ROCm repository and install tools
$ wget https://repo.radeon.com/rocm/rocm.gpg.key -O - | gpg --dearmor | sudo tee /etc/apt/keyrings/rocm.gpg > /dev/null
$ echo "deb [arch=amd64 signed-by=/etc/apt/keyrings/rocm.gpg] https://repo.radeon.com/rocm/apt/latest noble main" | sudo tee /etc/apt/sources.list.d/rocm.list
$ sudo apt update
$ sudo apt install rocm-dev rocm-libs -y
# Python environment setup (using standard 3.12 from Ubuntu 24.04)
$ sudo apt install python3.12-venv python3.12-dev
$ mkdir -p ~/ComfyUI && cd ~/ComfyUI
$ python3.12 -m venv venv_rocm711
$ source venv_rocm711/bin/activate
# [Important] Installing TheROCk 7.11 (Nightly)
# Install by specifying the index dedicated to Strix Halo (gfx1151)
$ pip install --index-url https://rocm.nightlies.amd.com/v2/gfx1151/ \
torch torchvision torchaudio --force-reinstall
4.3 Adding Missing Libraries and Verifying Operation
Since some libraries are missing when running rocminfo, we will perform additional installations.
# Install dependent libraries
$ sudo apt update
$ sudo apt install libatomic1 libnuma1 -y
$ sudo apt install libgomp1 libgl1 libglx-mesa0 libglib2.0-0 -y
Run rocminfo and confirm that the GPU (gfx1151) is correctly recognized.
$ rocminfo
ROCk module is loaded
=====================
HSA System Attributes
=====================
Runtime Version: 1.18
Runtime Ext Version: 1.15
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+ 395 w/ Radeon 8060S
Uuid: CPU-XX
Marketing Name: AMD RYZEN AI MAX+ 395 w/ Radeon 8060S
Vendor Name: CPU
Feature: None specified
Profile: FULL_PROFILE
Float Round Mode: NEAR
Max Queue Number: 0(0x0)
Queue Min Size: 0(0x0)
Queue Max Size: 0(0x0)
Queue Type: MULTI
Node: 0
Device Type: CPU
Cache Info:
L1: 49152(0xc000) KB
Chip ID: 0(0x0)
ASIC Revision: 0(0x0)
Cacheline Size: 64(0x40)
Max Clock Freq. (MHz): 5187
BDFID: 0
Internal Node ID: 0
Compute Unit: 32
SIMDs per CU: 0
Shader Engines: 0
Shader Arrs. per Eng.: 0
WatchPts on Addr. Ranges:1
Memory Properties:
Features: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: FINE GRAINED
Size: 128135472(0x7a33130) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 128135472(0x7a33130) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 3
Segment: GLOBAL; FLAGS: KERNARG, FINE GRAINED
Size: 128135472(0x7a33130) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
Pool 4
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 128135472(0x7a33130) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:4KB
Alloc Alignment: 4KB
Accessible by all: TRUE
ISA Info:
*******
Agent 2
*******
Name: gfx1151
Uuid: GPU-XX
Marketing Name: Radeon 8060S Graphics
Vendor Name: AMD
Feature: KERNEL_DISPATCH
Profile: BASE_PROFILE
Float Round Mode: NEAR
Max Queue Number: 128(0x80)
Queue Min Size: 64(0x40)
Queue Max Size: 131072(0x20000)
Queue Type: MULTI
Node: 1
Device Type: GPU
Cache Info:
L1: 32(0x20) KB
L2: 2048(0x800) KB
L3: 32768(0x8000) KB
Chip ID: 5510(0x1586)
ASIC Revision: 0(0x0)
Cacheline Size: 128(0x80)
Max Clock Freq. (MHz): 2900
BDFID: 48640
Internal Node ID: 1
Compute Unit: 40
SIMDs per CU: 2
Shader Engines: 2
Shader Arrs. per Eng.: 2
WatchPts on Addr. Ranges:4
Coherent Host Access: FALSE
Memory Properties: APU
Features: KERNEL_DISPATCH
Fast F16 Operation: TRUE
Wavefront Size: 32(0x20)
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Max Waves Per CU: 32(0x20)
Max Work-item Per CU: 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
Max fbarriers/Workgrp: 32
Packet Processor uCode:: 32
SDMA engine uCode:: 17
IOMMU Support:: None
Pool Info:
Pool 1
Segment: GLOBAL; FLAGS: COARSE GRAINED
Size: 130023424(0x7c00000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 2
Segment: GLOBAL; FLAGS: EXTENDED FINE GRAINED
Size: 130023424(0x7c00000) KB
Allocatable: TRUE
Alloc Granule: 4KB
Alloc Recommended Granule:2048KB
Alloc Alignment: 4KB
Accessible by all: FALSE
Pool 3
Segment: GROUP
Size: 64(0x40) KB
Allocatable: FALSE
Alloc Granule: 0KB
Alloc Recommended Granule:0KB
Alloc Alignment: 0KB
Accessible by all: FALSE
ISA Info:
ISA 1
Name: amdgcn-amd-amdhsa--gfx1151
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
ISA 2
Name: amdgcn-amd-amdhsa--gfx11-generic
Machine Models: HSA_MACHINE_MODEL_LARGE
Profiles: HSA_PROFILE_BASE
Default Rounding Mode: NEAR
Default Rounding Mode: NEAR
Fast f16: TRUE
Workgroup Max Size: 1024(0x400)
Workgroup Max Size per Dimension:
x 1024(0x400)
y 1024(0x400)
z 1024(0x400)
Grid Max Size: 4294967295(0xffffffff)
Grid Max Size per Dimension:
x 2147483647(0x7fffffff)
y 65535(0xffff)
z 65535(0xffff)
FBarrier Max Size: 32
*** Done ***
5. ComfyUI Setup
Install the ComfyUI core and its dependent packages.
$ cd ~/ComfyUI
$ pip install -r requirements.txt
$ pip install -r manager-requirements.txt
$ pip install deepdiff gguf aiohttp_sse piexif
# Suppress Git warnings
$ export GIT_PYTHON_REFRESH=quiet
Verifying Startup:
# Startup command (Manager enabled, Legacy UI)
$ python main.py --listen --enable-manager --enable-manager-legacy-ui
Access http://hostname.local:8188 from any browser to confirm that ComfyUI is running.
6. Automatic Startup Configuration (Systemd)
This configuration automatically starts ComfyUI within Distrobox when the PC boots up.
Create ~/.config/systemd/user/comfyui.service.
$ mkdir -p ~/.config/systemd/user
$ vi ~/.config/systemd/user/comfyui.service
Contents of comfyui.service:
[Unit]
Description=ComfyUI inside Distrobox (strix-ai)
After=network.target
[Service]
Type=simple
# By using bash -i, the .bashrc settings are loaded and environment variables are applied
ExecStart=/usr/bin/distrobox-enter -n strix-ai -- bash -i -c "source ~/ComfyUI/venv_rocm711/bin/activate && python ~/ComfyUI/main.py --listen --enable-manager --enable-manager-legacy-ui"
Restart=always
RestartSec=10
WorkingDirectory=%h/ComfyUI
[Install]
WantedBy=default.target
Enable and start the service.
# Reload configuration
$ systemctl --user daemon-reload
# Enable automatic startup
$ systemctl --user enable comfyui.service
# Start immediately
$ systemctl --user start comfyui.service
Discussion