No NPU Usage

[cpreisler]

I'm not seeing any NPU usage in Resources. I'm running an LLM on FastFlowLM. I ran a long running task on the LLM. I'm still not seeing any NPU usage. Why is Resources not showing any NPU usage?

I'm running Resource 1.10.2.

Here is the LLM I'm running. If you need any additional information let me know.

chad@tuxedoinfinity:~/AI$ flm run llama3.1:8b
[FLM]  Loading model: /home/chad/.config/flm/models/Llama-3.1-8B-NPU2
[FLM]  Type /? for help
>>> How are you today?

I'm just a language model, so I don't have emotions or feelings like humans do.

Here is some information about my system.

chad@tuxedoinfinity:~$ flm validate
[Linux]  Kernel: 6.17.0-108014-tuxedo
[Linux]  NPU: /dev/accel/accel0 with 8 columns
[Linux]  NPU FW Version: 1.1.2.64
[Linux]  amdxdna version: 0.6
[Linux]  Memlock Limit: infinity
chad@tuxedoinfinity:~$ lsb_release -a
No LSB modules are available.
Distributor ID: Tuxedo
Description:    TUXEDO OS
Release:        24.04
Codename:       noble

chad@tuxedoinfinity:~$ xrt-smi examine
181 (127753427937216): PID(126439): Created pcidev (0000:66:00.1)
310644 (127753427937216): PID(126439): Opened /dev/accel/accel0 as 3, sysfs: /sys/bus/pci/devices/0000:66:00.1
334379 (127753427937216): PID(126439): Expanding BO from 0 to 67108864
9899702 (127753427937216): PID(126439): Created AMDXDNA_BO_DEV_HEAP: type= hdl=1 sz=0x4000000 paddr=0x4000000 vaddr=0x7430e4000000 uptr=0x0
9914159 (127753427937216): PID(126439): Created device (0000:66:00.1) ...
9963091 (127753427937216): PID(126439): query key (174) is not supported
10057759 (127753427937216): PID(126439): query key (22) is not supported
133390335 (127753427937216): PID(126439): query key (174) is not supported
133494871 (127753427937216): PID(126439): query key (22) is not supported
134733609 (127753427937216): PID(126439): query key (174) is not supported
134763785 (127753427937216): PID(126439): query key (22) is not supported
System Configuration
  OS Name              : Linux
  Release              : 6.17.0-108014-tuxedo
  Machine              : x86_64
  CPU Cores            : 20
  Memory               : 92135 MB
  Distribution         : TUXEDO OS
  GLIBC                : 2.39
  Model                : TUXEDO InfinityBook Pro AMD Gen10
  BIOS Vendor          : American Megatrends International, LLC.
  BIOS Version         : N.1.20A13
  Processor            : AMD Ryzen AI 9 365 w/ Radeon 880M

XRT
  Version              : 2.21.0
  Branch               : HEAD
  Hash                 : a07ada1c2cb33b79d8d3a96efe853b2452e57711
  Hash Date            : 2025-12-14 11:26:53
  amdxdna              : unknown, unknown
  virtio-pci           : unknown, unknown
  NPU Firmware Version : 1.1.2.64

Device(s) Present
|BDF             |Name          |
|----------------|--------------|
|[0000:66:00.1]  |RyzenAI-npu4  |


135007443 (127753427937216): PID(126439): Destroying device (0000:66:00.1) ...
135019837 (127753427937216): PID(126439): Destroying AMDXDNA_BO_DEV_HEAP: type= hdl=1 sz=0x4000000 paddr=0x4000000 vaddr=0x7430e4000000 uptr=0x0
139479857 (127753427937216): PID(126439): Closed 3
139492050 (127753427937216): PID(126439): Destroying pcidev (0000:66:00.1)
chad@tuxedoinfinity:~$ 

chad@tuxedoinfinity:~$ sudo dmesg | grep -i xdna
[sudo] password for chad: 
[   22.363860] amdxdna 0000:66:00.1: [drm] Load firmware amdnpu/17f0_10/npu_7.sbin
[   22.363868] amdxdna 0000:66:00.1: enabling device (0000 -> 0002)
[   22.513015] [drm] Initialized amdxdna_accel_driver 0.6.0 for 0000:66:00.1 on minor 0
[   25.260172]  videobuf2_v4l2 snd_rawmidi bluetooth snd_pci_acp6x ttm videobuf2_common uniwill_wmi(OE) mac80211 snd_pci_acp5x tuxedo_io(OE) drm_exec clevo_wmi(OE) kvm snd_seq snd_rn_pci_acp3x drm_suballoc_helper snd_pcm tuxedo_keyboard(OE) amd_pmf snd_seq_device snd_acp_config videodev tuxedo_compatibility_check(OE) drm_display_helper asus_wmi amdtee led_class_multicolor snd_soc_acpi snd_timer irqbypass cec rapl amdxdna(OE) 

[cpreisler]

I'm going to answer my own question. The Ryzen AI 9 365 NPU was correctly identified and partially operational. To fix it I had to
ensure the XRT Shim (plugin) was correctly installed and linked to my kernel driver. So I built and installed the base and shim from source and rebooted. Now Resources is showing NPU usage when I run an LLM on FastFlowLM. Yay!