Ubuntu release
lsb_release -a 2>/dev/null
returns
Distributor ID: Ubuntu
Description: Ubuntu 26.04 LTS
Release: 26.04
Codename: resolute
uname -r
returns
7.0.0-22-generic
GPU setup
The Framework 16 with AMD Ryzen CPU ships with an integrated GPU (iGPU) and a dedicated Neural Processing Unit (NPU). Moreover the AMD RX 7700S provides a dedicated GPU (dGPU) graphics card.
Kernel modules
sudo lsmod |grep -iE 'amd(gpu|xdna)'
returns
amdxdna 172032 0
amd_pmf 131072 1 amdxdna
amdgpu 21549056 55
amdxcp 12288 1 amdgpu
i2c_algo_bit 16384 1 amdgpu
drm_ttm_helper 20480 2 amdgpu
ttm 135168 2 amdgpu,drm_ttm_helper
drm_exec 12288 1 amdgpu
drm_panel_backlight_quirks 12288 1 amdgpu
gpu_sched 69632 2 amdxdna,amdgpu
drm_suballoc_helper 24576 1 amdgpu
drm_buddy 28672 1 amdgpu
drm_display_helper 303104 1 amdgpu
video 77824 1 amdgpu
cec 106496 2 drm_display_helper,amdgpu
More details on the kernel modules utilized:
modinfo amdgpu | grep -E "version:|description:|vermagic:|filename:"
returns
filename: /lib/modules/7.0.0-22-generic/kernel/drivers/gpu/drm/amd/amdgpu/amdgpu.ko.zst
description: AMD GPU
srcversion: 3A899DD7C85BDF64918FEF2
vermagic: 7.0.0-22-generic SMP preempt mod_unload modversions
and
modinfo amdxdna | grep -E "version:|description:|vermagic:|filename:"
returns
filename: /lib/modules/7.0.0-22-generic/kernel/drivers/accel/amdxdna/amdxdna.ko.zst
description: amdxdna driver
srcversion: 3873C4BB9348F32F5C9E515
vermagic: 7.0.0-22-generic SMP preempt mod_unload modversions
AMD base driver
amd-smi
+------------------------------------------------------------------------------+
| AMD-SMI 26.2.1+unknown amdgpu version: 7.0.0-22 ROCm version: 7.2.1 |
| VBIOS version: 00079185 |
| Platform: Linux Baremetal |
|-------------------------------------+----------------------------------------|
| BDF GPU-Name | Mem-Uti Temp UEC Power-Usage |
| GPU HIP-ID OAM-ID Partition-Mode | GFX-Uti Fan Mem-Usage |
|=====================================+========================================|
| 0000:c4:00.0 AMD Radeon RX 7700S | 0 % 39 °C 0 2/100 W |
| 0 0 N/A N/A | 0 % 29.8 % 26/8176 MB |
|-------------------------------------+----------------------------------------|
| 0000:c5:00.0 ...adeon 890M Graphics | N/A N/A 0 N/A |
| 1 1 N/A N/A | N/A N/A 471/512 MB |
+-------------------------------------+----------------------------------------+
+------------------------------------------------------------------------------+
| Processes: |
| GPU PID Process Name GTT_MEM VRAM_MEM MEM_USAGE CU % |
|==============================================================================|
| No running processes found |
+------------------------------------------------------------------------------+
AMD ROCm framework
rocminfo | grep -E "Name:|Compute Unit|Max Clock|Memory Size|Agent"
HSA Agents
Agent 1
Name: AMD Ryzen AI 9 HX 370 w/ Radeon 890M
Marketing Name: AMD Ryzen AI 9 HX 370 w/ Radeon 890M
Vendor Name: CPU
Max Clock Freq. (MHz): 2605
Compute Unit: 24
Agent 2
Name: gfx1100
Marketing Name: AMD Radeon RX 7700S
Vendor Name: AMD
Max Clock Freq. (MHz): 2208
Compute Unit: 32
Name: amdgcn-amd-amdhsa--gfx1100
Name: amdgcn-amd-amdhsa--gfx11-generic
Agent 3
Name: gfx1100
Marketing Name: AMD Radeon 890M Graphics
Vendor Name: AMD
Max Clock Freq. (MHz): 2900
Compute Unit: 16
Name: amdgcn-amd-amdhsa--gfx1100
Name: amdgcn-amd-amdhsa--gfx11-generic
Agent 4
Name: aie2
Marketing Name: AIE-ML
Vendor Name: AMD
Max Clock Freq. (MHz): 0
Compute Unit: 0
OpenCL framework
grep -H '' /etc/OpenCL/vendors/*
returns
/etc/OpenCL/vendors/amdocl64.icd:libamdocl64.so
/etc/OpenCL/vendors/amdxrt.icd:/usr/./lib/x86_64-linux-gnu/libxilinxopencl.so
/etc/OpenCL/vendors/rusticl.icd:libRusticlOpenCL.so.1
and
sudo clinfo | grep -E "Number of devices|Number of platforms|Device Name|Device Type|Platform Name|Max compute units|Max clock|Global memory size"
returns
Number of platforms 2
Platform Name AMD Accelerated Parallel Processing
Platform Name Xilinx
Platform Name AMD Accelerated Parallel Processing
Number of devices 2
Device Name gfx1102
Device Type GPU
Max compute units 16
Max clock frequency 2208MHz
Global memory size 8573157376 (7.984GiB)
Device Name gfx1150
Device Type GPU
Max compute units 8
Max clock frequency 2900MHz
Global memory size 32794570752 (30.54GiB)
Platform Name Xilinx
Number of devices 1
XRT build version: 2.21.75
PID: 370347
UID: 0
[Sat Jun 13 08:36:36 2026 GMT]
HOST: Device Name <printDeviceInfo:0: get CL_DEVICE_NAME size : error -6>
Device Type Accelerator
Max compute units 0
Max clock frequency 0MHz
cma-framework
EXE: /usr/bin/clinfo
[XRT] ERROR: get_device_info: Operation not supported
[XRT] ERROR: clGetDeviceInfo: invalid param_name
Global memory size 0
Platform Name AMD Accelerated Parallel Processing
Device Name gfx1102
Platform Name AMD Accelerated Parallel Processing
Device Name gfx1102
Device Name gfx1150
Platform Name AMD Accelerated Parallel Processing
Device Name gfx1102
Device Name gfx1150
LLM inference through Lemonade
sudo apt install lemonade-{server,desktop}
lemonade --version && lemond --version
yields
lemonade version 10.7.0
lemond version 10.7.0
and successful integration and usage of the NPU with tools like https://aider.chat running
OPENAI_API_BASE='http://localhost:13305/v1' \
OPENAI_API_KEY='' \
aider --model openai/llama3.2-1b-FLM
Scientific Computing with i/dGPUs in Python
- PyTorch: https://pytorch.org
- CuPy: https://cupy.dev
Utilizing a micromamba environment and pip installs:
micromamba list -n framework | grep -iE 'torch|cupy'
returns
cupy-rocm-7-0 14.1.1 pypi_0 pypi
torch 2.10.0+rocm7.0 pypi_0 pypi
torchaudio 2.10.0+rocm7.0 pypi_0 pypi
torchvision 0.25.0+rocm7.0 pypi_0 pypi
installed through
pip install torch{,vision,audio} --index-url https://download.pytorch.org/whl/rocm7.0
pip install cupy-rocm-7-0
Both GPU’s are accessible:
# PyTorch
import torch
torch.tensor([.0]).to('cuda:0').device
torch.tensor([1.]).to('cuda:1').device
# CuPy
import cupy
with cupy.cuda.Device(0):
cupy.array([.0]).device
with cupy.cuda.Device(1):
cupy.array([1.]).device