🖥️ Why EndeavourOS

• I initially tried Kubuntu for several reasons, but I ran into persistent issues with my Wi-Fi driver.

• Eventually, I decided to give EndeavourOS a try. To my surprise, everything ran much better,

• and I found Arch Linux to be just as approachable as Ubuntu. After spending hours researching and experimenting,

• I put together this short guide to document how I successfully enabled GPU acceleration for PixInsight.

🖥️ System Environment

• OS: EndeavourOS (Arch Linux)

• CPU: 32 × 13th Gen Intel® Core™ i9-13900K

• GPU: NVIDIA GeForce RTX 4070 Ti/PCIe/SSE2

• Driver: NVIDIA 580.105.08

• PixInsight Location: /opt/PixInsight/bin/

1. Install CUDA Toolkit 12.5.1

1. Download the local run file:

   wget https://developer.download.nvidia.com/compute/cuda/12.5.1/local_installers/cuda_12.5.1_555.42.06_linux.run

2. Make it executable (if needed):

   chmod +x cuda_12.5.1_555.42.06_linux.run

3. Install toolkit only (no driver, no gcc13):

   sudo ./cuda_12.5.1_555.42.06_linux.run --silent --toolkit --toolkitpath=/usr/local/cuda-12.5.1 --override

2. Install AUR Helper (yay)

1. sudo pacman -S --needed git base-devel
   git clone https://aur.archlinux.org/yay.git
   cd yay
   makepkg -si

3. Install libxml2-legacy

1. Using yay:

   yay -S libxml2-legacy

4. Install cuDNN 9.3.0

1. Download tarball:

   wget https://developer.download.nvidia.com/compute/cudnn/redist/cudnn/linux-x86_64/cudnn-linux-x86_64-9.3.0.75_cuda12-archive.tar.xz

2. Unpack:

   tar -xvf cudnn-linux-x86_64-9.3.0.75_cuda12-archive.tar.xz

3. Copy headers and libraries:

   sudo cp cudnn-linux-x86_64-9.3.0.75_cuda12-archive/include/cudnn*.h /usr/local/cuda-12.5.1/include
   sudo cp -P cudnn-linux-x86_64-9.3.0.75_cuda12-archive/lib/libcudnn* /usr/local/cuda-12.5.1/lib64
   sudo chmod a+r /usr/local/cuda-12.5.1/include/cudnn*.h /usr/local/cuda-12.5.1/lib64/libcudnn*

5. Installing TensorFlow C Library on Linux

1. Download the archive Save the libtensorflow-gpu-linux-x86_64.tar.gz file to your home directory (/home/$USER/).
   You can download it directly from the official TensorFlow site: https://www.tensorflow.org/install/lang_c

   TensorFlow C installation guide.
   Create a directory for TensorFlow
   bash
   mkdir ~/tensorflow
   Move the archive into the new directory
   bash
   mv ~/libtensorflow-gpu-linux-x86_64.tar.gz ~/tensorflow/
   Navigate into the directory
   bash
   cd ~/tensorflow
   Extract the archive
   bash
   tar -xvf libtensorflow-gpu-linux-x86_64.tar.gz

---

6. Configure PixInsight

1. Locate the startup script:

   /opt/PixInsight/bin/PixInsight.sh

2. Edit the line starting with LD_LIBRARY_PATH (usually line 7). Update it to:

   LD_LIBRARY_PATH=$HOME/tensorflow/lib:/usr/local/cuda-12.5.1:$dirname/lib:$dirname

3. Save the file.

---

✅ Result

• TensorFlow 2.18 (prebuilt) works seamlessly with CUDA 12.5.1 + cuDNN 9.3.