🚀 Full Guide: Installing NVIDIA RTX 3050, CUDA 12.8 & cuDNN 8.9.7 on Ubuntu 22.04

This guide ensures your system is ready for deep learning, GPU acceleration, and high-performance computing using the NVIDIA RTX 3050.

---

📋 Prerequisites

• Ubuntu 22.04 (or newer)

• RTX 3050 GPU

• Internet connection

• Terminal access (sudo)

• Charger plugged in (for laptops)

---

✅ STEP 1: Update Your System

sudo apt update && sudo apt upgrade -y

🔍 Why?
To ensure all existing packages are up to date, reducing the chance of version conflicts during driver or CUDA installation.
👉 Think of it like cleaning your desk before setting up a new PC build.

Optional (if prompted or just to be safe):

sudo reboot

---

🚫 STEP 2: Disable the Nouveau Driver (Default Open-Source NVIDIA Driver)

Ubuntu uses Nouveau by default, but it conflicts with NVIDIA’s proprietary driver, so we need to blacklist it.

2.1 Create a Blacklist File

sudo nano /etc/modprobe.d/blacklist-nouveau.conf

Paste this into the file:

blacklist nouveau
options nouveau modeset=0

Save and exit:

• Ctrl + O, Enter → Save

• Ctrl + X → Exit

2.2 Update initramfs

sudo update-initramfs -u

🔍 Why?
Rebuilds the initial RAM filesystem with the new kernel module settings, so Nouveau stays disabled.

2.3 Reboot

sudo reboot

---

📦 STEP 3: Install the Official NVIDIA Driver

3.1 Check Available Drivers

ubuntu-drivers devices

🔍 Why?
Shows the best driver recommended for your GPU (e.g., nvidia-driver-570). You avoid guesswork.

3.2 Install Recommended Driver

sudo apt install nvidia-driver-570 -y

3.3 Reboot Again

sudo reboot

🔍 Why?
So the system starts using the new NVIDIA driver instead of the disabled Nouveau.

3.4 Verify GPU is Working

nvidia-smi

You should see output like this:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 570.xx    Driver Version: 570.xx    CUDA Version: 12.x          |
| GPU       Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr.|
| 0     NVIDIA RTX 3050      Off      | 00000000:01:00.0  On |                  |
+-----------------------------------------------------------------------------+

Congrats! You're running NVIDIA RTX 3050 successfully on Ubuntu 🎉

🔍 Why?
Confirms that your RTX 3050 is detected and functioning correctly with the installed driver.

---

💻 STEP 4: Install CUDA Toolkit 12.8

You need this for GPU-based computations in ML, AI, and data science tasks.

4.1 Download .deb (local) from the official CUDA site

Then install it:

sudo dpkg -i cuda-repo-ubuntu2204-12-8-local_*.deb
sudo cp /var/cuda-repo-*/cuda-*-keyring.gpg /usr/share/keyrings/
sudo apt update
sudo apt install cuda -y

🔍 Why?
Adds the NVIDIA CUDA repo and installs the CUDA dev kit — including compiler (nvcc), runtime libraries, and more.

4.2 Add CUDA to Path

Open .bashrc:

nano ~/.bashrc

Add these lines at the bottom:

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

Then apply the changes:

source ~/.bashrc

🔍 Why?
So your terminal knows where to find CUDA tools and libraries.

4.3 Verify CUDA

nvcc --version

🔍 Why?
To confirm CUDA installed correctly and the version is available.

---

⚙️ STEP 5: Install cuDNN 8.9.7 (for Deep Learning Acceleration)

cuDNN boosts ML frameworks like PyTorch & TensorFlow by giving GPU-based neural net performance.

5.1 Download .deb from cuDNN Archive

Install it:

sudo dpkg -i cudnn-local-repo-ubuntu2204-8.9.7.29_1.0-1_amd64.deb
sudo cp /var/cudnn-local-repo-*/cudnn-*-keyring.gpg /usr/share/keyrings/
sudo apt update
sudo apt install libcudnn8 libcudnn8-dev libcudnn8-samples -y

🔍 Why?

• libcudnn8: Runtime for deep learning

• libcudnn8-dev: Headers for compiling models

• libcudnn8-samples: Sample models to test

---

🧪 STEP 6: Test cuDNN Installation

cat /usr/include/cudnn_version.h | grep CUDNN_MAJOR -A 2

🔍 Why?
Checks the installed cuDNN version.

Run a sample test:

cp -r /usr/src/cudnn_samples_v8/ ~/cudnn_samples
cd ~/cudnn_samples/mnistCUDNN
make
./mnistCUDNN

✅ Output should say: Test Passed!
🔍 Why?
Confirms everything (CUDA + cuDNN) is correctly set up.

---

🧠 STEP 7: Verify GPU with Python (PyTorch / TensorFlow)

For PyTorch:

import torch
print(torch.cuda.is_available())        # Should return True
print(torch.cuda.get_device_name(0))    # NVIDIA RTX 3050

For TensorFlow:

from tensorflow.python.client import device_lib
print(device_lib.list_local_devices())

🔍 Why?
To confirm your ML code can now access the GPU for training or inference.

---

🔌 Pro Tip: Keep Charger Plugged In

Laptop GPUs like RTX 3050 may disable themselves on battery to save power.
👉 Always keep the charger connected during GPU-intensive tasks.

---

🎉 That’s it!

You now have a complete setup with:

• ✅ Official NVIDIA Driver

• ✅ CUDA 12.8

• ✅ cuDNN 8.9.7

• ✅ Fully GPU-enabled deep learning environment

---

Made with 🫶🏻 by Ronnie Allen