Enable NPU on Board | Radxa Docs

[giscus[bot]]

Enable NPU on Board | Radxa Docs

When using system image version R2 or higher, the NPU runtime environment is pre-installed and no additional installation is required.

https://docs.radxa.com/en/dragon/q6a/app-dev/npu-dev/fastrpc-setup

[cupuyc1]

My tests are FAILED... what im doing wrong?
cup@radxa-dragon-q6a:$ sudo apt install fastrpc libcdsprpc1
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
fastrpc is already the newest version (1.0.3-1).
fastrpc set to manually installed.
libcdsprpc1 is already the newest version (1.0.3-1).
libcdsprpc1 set to manually installed.
0 upgraded, 0 newly installed, 0 to remove and 0 not upgraded.
cup@radxa-dragon-q6a:$ sudo apt install fastrpc-test
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following NEW packages will be installed:
fastrpc-test
0 upgraded, 1 newly installed, 0 to remove and 0 not upgraded.
Need to get 42.4 kB of archives.
After this operation, 402 kB of additional disk space will be used.
Get:1 https://radxa-repo.github.io/qcs6490-noble qcs6490-noble/main arm64 fastrpc-test arm64 1.0.3-1 [42.4 kB]
Fetched 42.4 kB in 0s (120 kB/s)
Selecting previously unselected package fastrpc-test.
(Reading database ... 171343 files and directories currently installed.)
Preparing to unpack .../fastrpc-test_1.0.3-1_arm64.deb ...
Unpacking fastrpc-test (1.0.3-1) ...
Setting up fastrpc-test (1.0.3-1) ...
Scanning processes...
Scanning processor microcode...
Scanning linux images...

Running kernel seems to be up-to-date.

The processor microcode seems to be up-to-date.

No services need to be restarted.

No containers need to be restarted.

No user sessions are running outdated binaries.

No VM guests are running outdated hypervisor (qemu) binaries on this host.
cup@radxa-dragon-q6a:~$ fastrpc_test -a v68

Allocate 4000 bytes from ION heap
ERROR 0x68: memory alloc failed
[FAIL] libcalculator.so (error code: 0x68)

ERROR 0x72: Unable to create FastRPC session on domain 3
Exiting...
Test FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
[FAIL] libmultithreading.so (error code: 0x72)

ERROR 0x72: Unable to create FastRPC session on domain 3

hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector FARF_RUNTIME (id: 0): 114
ERROR 0x72: Unable to create FastRPC session on domain 3

hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector MEM (id: 1): 114
ERROR 0x72: Unable to create FastRPC session on domain 3

hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector PERF (id: 2): 114
[FAIL] libhap_example.so (error code: 0x72)

========================================
Test Summary:
Total tests run: 3
Passed: 0
Failed: 3
Skipped: 0

RESULT: 3 test(s) FAILED
cup@radxa-dragon-q6a:~$

[tangzz-radxa]

The test failures (error 0x68 and 0x72) indicate the DSP domain 3 (CDSP) is not accessible. Error 0x72 means the FastRPC session cannot be created on domain 3, and error 0x68 follows as a consequence of the session failure.

Please check the following:

1. Verify OS image version
Error 0x72 typically occurs on older images where the NPU firmware is not loaded. Make sure you are running RadxaOS R2 or higher:

cat /etc/radxa-os-version
# or
cat /etc/os-release

2. Check FastRPC device nodes

ls -la /dev/fastrpc-*

You should see: , , . If any are missing, the DSP driver is not loaded.

3. Check DSP firmware status

# Check if the DSP firmware is loaded
cat /sys/kernel/debug/remoteproc/remoteproc*/name 2>/dev/null
# Or check for any DSP-related kernel messages
dmesg | grep -i dsp
dmesg | grep -i fastrpc

4. Check if the NPU kernel driver is loaded

lsmod | grep -i qcom
lsmod | grep -i npu
dmesg | grep -i npu

If is missing or shows no output, the DSP driver/firmware is not loaded. On RadxaOS R2+, the firmware should load automatically at boot. If you are on an older image, you may need to reflash with the latest RadxaOS image.

If the issue persists after confirming R2+ image and all device nodes present, please share the output of the commands above — that will help narrow down whether it is a firmware loading issue or a permission issue.

Reference: https://docs.radxa.com/en/dragon/q6a/app-dev/npu-dev/fastrpc-setup

[cupuyc1]

ok, it looks like without "sudo" that test (fastrpc_test -a v68) doesnt work ... or another way: need to add current user to the "render" group.
Please update the "Enable NPU on Board" - documentation page.