- NUC12WSHi5
- PopOS 22.04 nvidia
- Razer Core X
- RTX 3060 12GB
- enter F2 in boot
- disable Secure Boot
- enable 3rd UEFI Driver
use boltctl
to controll TB3 in Linux
boltctl auth
boltctl enroll
nvidia-smi
(nvidia-smi
's cuda version is max version by GPU and not current system version)
- RTX 3060 -> Compute Capability 8.6
- latest by PopOS
- latest by PopOS is 11.2
- CUDA 11.2 -> Compute Capability 5.0 ~ 8.7
sudo apt-get install system76-cuda-11.2 system76-cudnn-11.2
(cudnn package version suffix is based on CUDA version as system76-cudnn-${CUDA_VERSION}
)
- Support Matrix :: NVIDIA Deep Learning cuDNN Documentation
- CUDA and cuDNN (Install) - System76 Support
- Intel NUC + SSD 2TB + DRAM 32GBでハイスペックLinuxデスクトップを作ってみた - Qiita
- eGPUでハイスペックLinuxデスクトップをDeep Learning Workstation化計画(eGPUセットアップ編) - Qiita
- eGPU+Ubuntu20.04LTSを使って、機械学習環境(TensorFlow, PyTorch)を作成してみた。 - Qiita