Linux(Ubuntu/Debian)에 NVIDIA GPU 드라이버 설치

공식 페이지

NVIDIA GPU 드라이버 설치

Ubuntu/Debian에서 자동 업데이트 비활성화

# 1. Disable auto-update and auto-upgrade:
echo 'APT::Periodic::Update-Package-Lists "0";' | sudo tee /etc/apt/apt.conf.d/20auto-upgrades
echo 'APT::Periodic::Unattended-Upgrade "0";' | sudo tee -a /etc/apt/apt.conf.d/20auto-upgrades

오픈 소스 nouveau 드라이버 비활성화

# 2. Write blacklist options for the nouveau driver (disable):
echo "blacklist nouveau" | sudo tee -a /etc/modprobe.d/blacklist-nouveau.conf
echo "options nouveau modeset=0" | sudo tee -a /etc/modprobe.d/blacklist-nouveau.conf

# 3. Update kernel initramfs for blacklist (modprobe.d) options:
sudo update-initramfs -u

필수 패키지 설치

# 4. Clean apt package manager lists and caches:
sudo rm -vrf /var/lib/apt/lists/* /var/cache/apt/archives/* /tmp/*
sudo apt-get clean -y

# 5. Updating, indexing, and caching apt package manager app lists:
sudo apt-get update --fix-missing -o Acquire::CompressionTypes::Order::=gz

# 6. Upgrade and reinstall OS base packages:
sudo apt-get upgrade -y

# 7. Remove old or unnecessary leftover packages:
sudo apt-get autoremove -y

# 8. IMPORTANT! Install dkms and base packages:
sudo apt-get install -y build-essential pkg-config libglvnd-dev mesa-utils wget dkms

현재 NVIDIA GPU 모델에 적합한 Linux 드라이버를 다운로드합니다.

mkdir -v ~/Downloads
cd ~/Downloads

# 9. (Skip this, if you've already downloaded) Download appropriate Linux driver for current NVIDIA GPU model
wget https://us.download.nvidia.com/[GPU_MODEL_TYPE_URL]/NVIDIA-Linux-x86_64-[DRIVER_VERSION].run
# For example:
wget https://us.download.nvidia.com/XFree86/Linux-x86_64/510.54/NVIDIA-Linux-x86_64-510.54.run

NVIDIA GPU 드라이버를 설치하기 전 준비 단계

[중요] GUI 환경을 콘솔 환경으로 변경

# 10. IMPORTANT! Entering into virtual console environment:
sudo chvt 1
# Or:
sudo chvt 3

NVIDIA GPU 드라이버 설치를 준비하기 위해 GUI 프로세스를 중지합니다.

# 11. Stop GUI service (lightdm/gdm/gdm3/kdm/slim/etc..):
sudo service lightdm stop
# Or:
sudo service gdm stop
# Or:
sudo service gdm3 stop

# 12. Initialize and refresh console environment without X11 (GUI) processes:
sudo init 3

# Change directory to NVIDIA-Linux-x86_64-***.run driver downloaded directory. For example:
cd ~/Downloads

# 13. Change permission of NVIDIA-Linux-x86_64-***.run driver installer file to executable:
chmod +x ./NVIDIA-Linux-x86_64-[DRIVER_VERSION].run
# For example:
chmod +x ./NVIDIA-Linux-x86_64-510.54.run

NVIDIA driver.run 파일 설치

# 14. IMPORTANT! Install nvidia driver with appropriate arguments:
sudo ./NVIDIA-Linux-x86_64-[DRIVER_VERSION].run --no-x-check --no-cc-version-check --dkms
# For example:
sudo ./NVIDIA-Linux-x86_64-510.54.run --no-x-check --no-cc-version-check --dkms

# 15. Remove downloaded file:
rm -vf ./NVIDIA-Linux-x86_64-[DRIVER_VERSION].run
# For example:
rm -vf ./NVIDIA-Linux-x86_64-510.54.run

# 16. Reboot system:
sudo shutdown -r now

# 17. Check nvidia driver is installed:
nvidia-smi

설치 후

[ML/DL 서버 전용] nvidia-persistence 모드 데몬 설정

# Install git to clone:
sudo apt-get install -y ssh git

# Download nvidia-persistenced source code from github:
git clone https://github.com/NVIDIA/nvidia-persistenced.git

# Install nvidia-persistenced service daemon:
cd nvidia-persistenced/init
sudo ./install.sh

# Remove downloaded files:
cd ../.. && rm -rf nvidia-persistenced

# Check nvidia-persistence mode is ON:
nvidia-smi
# Or check nvidia-persistenced.service is running:
systemctl status nvidia-persistenced.service

설치 완료

[선택 사항] NVIDIA GPU 모니터링 도구 설치

🌟📡 엔비탑

# Install python3 and pip3 packages:
sudo apt-get install -y python3 python3-pip

# Install nvitop into system python3 packages:
sudo pip3 install nvitop

# Check nvitop:
nvitop -m full

📊 엔비탑

# Install cmake, ncurses and git:
sudo apt-get install -y cmake libncurses5-dev libncursesw5-dev

# Download nvtop source code from github:
git clone https://github.com/Syllo/nvtop.git
mkdir -vp nvtop/build && cd nvtop/build

cmake ..
# If it errors with "Could NOT find NVML (missing: NVML_INCLUDE_DIRS)"
# try the following command instead, otherwise skip to the build with make:
#cmake .. -DNVML_RETRIEVE_HEADER_ONLINE=True

make
sudo make install

# Remove downloaded files:
cd ../.. && rm -rf nvtop

# Check nvtop:
nvtop

NVIDIA 드라이버를 최신 버전으로 다시 설치하거나 업그레이드하십시오.

DKMS NVIDIA 모듈 제거

# Check current dkms nvidia modules, versions, and kernels:
# [TIP]: Use module versions to remove.
dkms status | grep nvidia

# Remove old dkms nvidia module/version:
sudo dkms remove nvidia/[DRIVER_VERSION] --all
# For example:
sudo dkms remove nvidia/510.54 --all

nvidia-persistenced 서비스 데몬 비활성화

# Stop and disable nvidia-persistenced service daemon:
sudo systemctl stop nvidia-persistenced.service
sudo systemctl disable nvidia-persistenced.service

모든 nvidia-* 패키지 제거

sudo apt-get remove --purge '^nvidia-.*'
# Or:
sudo nvidia-uninstall

# Reboot system:
sudo shutdown -r now

최신 NVIDIA 드라이버 버전 설치