GPU not detected on ZED Box Mini

Software
1

Hello, I am having an issue with the Zed Box Mini.
It seems the GPU or CUDA is not correctly detected
I don’t remember using a particular command that caused the issue, I just tried to install SDK 5.3 and then found the error. Don’t know if the installation caused this.

I am using jetpack 6.2.1, L4T 36.4.7

zed_diagnostic output

- Running ZED SDK Diagnostic : OK  

    ZED SDK version: 5.2.2

    CUDA version: V12.6.68

- Running Processor Diagnostic : OK  

    Processor:   ARMv8 Processor rev 1 (v8l)

    Motherboard:  Jetson, NVIDIA

Error: unable to open display stic : 50%

- Running Graphics Card Diagnostic : Failed

    Unable to detect your graphics card. Please install/repair (https://download.stereolabs.com/cuda12_web' style='color: #d9ff42;' >CUDA and update your NVIDIA graphics (https://download.stereolabs.com/nvidia/driver' style='color: #d9ff42;' >drivers.

- Running CUDA Operations Diagnostic : Failed

    CUDA Test: No CUDA-capable GPU found

Stack trace (most recent call last) in thread 3277:

#12   Object "[0xffffffffffffffff]", at 0xffffffffffffffff, in 

#11   Object "/usr/lib/aarch64-linux-gnu/libc.so.6", at 0xffffafaa9e9b, in 

#10   Object "/usr/lib/aarch64-linux-gnu/libc.so.6", at 0xffffafa40397, in 

#9    Object "/usr/lib/aarch64-linux-gnu/libstdc++.so.6", at 0xffffafd129cb, in 

#8    Object "ZED_Diagnostic", at 0x459b3f, in 

#7    Object "ZED_Diagnostic", at 0x46d527, in 

#6    Object "ZED_Diagnostic", at 0x4f060b, in 

#5    Object "/usr/local/zed/lib/libsl_ai.so", at 0xffff8f46cde7, in 

#4    Object "/usr/local/zed/lib/libsl_ai.so", at 0xffff8f45f227, in 

#3    Object "/usr/local/zed/lib/libsl_ai.so", at 0xffff8f45d8ab, in 

#2    Object "/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1", at 0xffffb28bc993, in 

#1    Object "/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1", at 0xffffb28a3c73, in 

#0    Object "/usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1", at 0xffffb29a64a0, in 

Segmentation fault (Address not mapped to object [0x8b])

Segmentation fault (core dumped)

This is the dmesg log generated by ZED_Diagnostic. The Json has not been generated.
dmesg_1.log (10.1 KB)

zed ros wrapper output

l::MAT_TYPE, sl::MEM) : Err [801]: operation not supported.

[component_container_isolated-2] in void sl::Mat::alloc(size_t, size_t, sl::MAT_TYPE, sl::MEM) : Err [801]: operation not supported.

[component_container_isolated-2] [2026-05-06 10:15:53 UTC][ZED][ERROR] No NVIDIA graphics card detected. Install an NVIDIA GPU, CUDA and restart your computer after completing installation.

[component_container_isolated-2] [2026-05-06 10:15:53 UTC][ZED][ERROR] NO GPU DETECTED in sl::ERROR_CODE sl::Camera::open(sl::InitParameters)

[component_container_isolated-2] [WARN] [1778062553.294415804] [zedxm.zed_node]: Error opening camera: NO GPU DETECTED

[component_container_isolated-2] [INFO] [1778062553.294579480] [zedxm.zed_node]: Please verify the camera connection

The Zed X driver was installed by using:
sudo dpkg -i --ignore-depends=nvidia-l4t-kernel "$TEMP_DIR/stereolabs-zedx-driver.deb"

I tried uninstalling CUDA and reinstalling by using the ZED SDK isntallation but I get the same error.

To uninstall CUDA I used:
sudo apt-get purge "cuda*" "libcudnn*" "libnvinfer*" -y
sudo apt-get autoremove --purge -y
sudo rm -rf /usr/local/cuda*

2

Hi @Alessio-Parmeggiani

This is not a good approach on Jetson because CUDA is normally installed with the Jetpack packages.

Please try this procedure to recover the device. If it does not work, then a flash to reset it is required.

Reinstall the BSP and JetPack components:

# Make sure apt sources are intact
cat /etc/apt/sources.list.d/nvidia-l4t-apt-source.list
# Should contain:
#   deb https://repo.download.nvidia.com/jetson/common r36.4 main
#   deb https://repo.download.nvidia.com/jetson/t234   r36.4 main

sudo apt update

# Reinstall the L4T GPU userspace driver (this is what was nuked)
sudo apt install --reinstall nvidia-l4t-cuda nvidia-l4t-3d-core nvidia-l4t-multimedia nvidia-l4t-core

# Pull the full JetPack stack back in (CUDA toolkit, cuDNN, TensorRT, VPI, etc.)
sudo apt install nvidia-jetpack

# Resolve any half-installed state from the original purge
sudo apt --fix-broken install -o Dpkg::Options::="--force-overwrite"

sudo reboot

After reboot:

ls -la /usr/lib/aarch64-linux-gnu/tegra/libcuda.so.1   # should resolve

Then reinstall the ZED SDK, and perform a test:

ZED_Diagnostic                     # should pass GPU + CUDA tests
3

Thanks for the answer! I followed your instructions but I get the same issue, I’ll try flashing again the zed box

1 Like
4

I completed the flash following the documentation but now I can’t see the wifi settings on the zed box, and I can’t connect to a wifi network.
There is something I can do or do I need an ethernet cable?

5

You must install the ZED X Driver package, which also contains the WiFi drivers.

You can download the installer with another PC and use a USB key to move it to the Jetson.

6

I installed the ZEDX Driver and did a reboot but still Wi-fi drivers seems to be not installed.

I also installed ZED SDK to check if at least the GPU issue solved but I can’t run ZED_Diagnostics because of this error:
Error while loading shared libraries: libqt5network.so.5: no such file or directory

7

I used an ethernet to install also ZED SDK and something else.
This is the output of ZED Diagnostics now (note that there is a ZEDX mini camera connected tot he ZED BOX mini) :

- Running ZED SDK Diagnostic : OK  
    ZED SDK version: 5.2.2
    CUDA version: V12.6.68
- Running Processor Diagnostic : OK  
    Processor:   ARMv8 Processor rev 1 (v8l)
    Motherboard:  Jetson, NVIDIA
Error: unable to open display stic : 50%
- Running Graphics Card Diagnostic : OK  
    Graphics card:  Orin
- Running CUDA Operations Diagnostic : OK  
    GPU 0 (Orin): CUDA operations working correctly
- Running AI Models Diagnostic : OK  
    MULTI CLASS DETECTION is not optimized
    MULTI CLASS MEDIUM DETECTION is not optimized
    MULTI CLASS ACCURATE DETECTION is not optimized
    HUMAN BODY FAST DETECTION is not optimized
    HUMAN BODY MEDIUM DETECTION is not optimized
    HUMAN BODY ACCURATE DETECTION is not optimized
    HUMAN BODY 38 FAST DETECTION is not optimized
    HUMAN BODY 38 MEDIUM DETECTION is not optimized
    HUMAN BODY 38 ACCURATE DETECTION is not optimized
    PERSON HEAD DETECTION is not optimized
    PERSON HEAD ACCURATE DETECTION is not optimized
    REID ASSOCIATION is not optimized
    NEURAL PLUS DEPTH is not optimized
    NEURAL LIGHT DEPTH is optimized
    NEURAL DEPTH is optimized
[2026-05-11 13:39:32.042] [ERROR] [SPSCDevice.cpp:494] Failed to connect to daemon socket: Connection refused
[2026-05-11 13:39:32.042] [ERROR] [SPSCDevice.cpp:118] Failed to set frequency for device 0
[2026-05-11 13:39:32.042] [ERROR] [IMUGMSLReader.cpp:2012] Failed to configure device 0
[2026-05-11 13:39:32.042] [ERROR] [IMUGMSLReader.cpp:235] Failed to initialize SPSC consumer
- Running USB Camera Diagnostic : Failed
    USB Camera not detected Flip the Type-C USB cable connected to the ZED or replug the camera. Make sure to use a USB 3.0 port.
- Running GMSL DRIVER Diagnostic : OK  
- Running Log Kernel Trace (for ZED X) Diagnostic : OK  

Diagnostic finished, a file has been generated : ZED_Diagnostic_Results.json

The GPU issue seems fixed but there are some other errors and the wifi is still not working after the ZED SDK and ZED X driver installation.

I tried: sudo apt install backport-iwlwifi-dkms -y and now the wifi started working again.

8

Hi @Alessio-Parmeggiani
I will share a script to install the WiFi driver as soon as possible.

Meanwhile, can you please share the content of the dmesg.log file generated by the command sudo ZED_Diagnostic --dmesg?
This will help me understand the problem that you have with the GMSL2 driver.

9

Nevermind, after a reboot the errors on the ZED_Diagnostics output disappeared

10

Hi @Alessio-Parmeggiani
This script should allow you to install the WiFi driver on the ZED Box Mini after the flash:
wifi_driver_install.sh (7.9 KB)

Please let me know if it works for you