1. Download the graphics card driver
Download the driver from the nvidia official website, select the device driver on the driver official website to search and download, pay attention to the corresponding operating system when searching,
generally install the NVIDIA Studio driver version, GeForce Game Ready is suitable for gamers, the following are two The official explanation of the version difference
The downloaded driver will be named in the form of version number. This version number will be used when installing cuda. The following is the downloaded driver file name 516.93-desktop
-win10-win11-64bit-international-nsd-dch-whql.exe
516.93 – graphics card driver version number
desktop – running platform
win10-win11-64bit – running operating system
nsd – nvidia studio version driver
dch – Declarative Componentized Hardware supported apps
whql – MicrosoftWindowsHardwareQualityLab operating system hardware quality laboratory logo
2. Install cuda
The installation of cuda needs to match the driver version of the graphics card. Currently, graphics cards above the 30 series only support cuda11 and above, so you need to pay attention when installing. Corresponding versions of
cuda and graphics card driver , the graphics card driver version displayed here is displayed in the form of version number. Select the corresponding version to download 3. Install cudnn cudnn download https://developer.nvidia.com/rdp/cudnn-download and compress the downloaded file to the corresponding directory of CUDA to overwrite it. 4. Tensorflow and CUDA, The version corresponding to CUDNN https://tensorflow.google.cn/install/source_windows#gpu
common connection
1. Download the graphics card driver
Download the graphics card driver
https://www.nvidia.cn/geforce/drivers/
2. Check the corresponding version
tensorrflow_cuda_cudnn
https://www.tensorflow.org/install/source_windows
torch_cuda_cudnn
https://pytorch.org/ get-started/previous-versions/
3. Download cuda, cudnn
cuda, cudnn, graphics driver corresponding version
https://docs.nvidia.cn/cuda/cuda-toolkit-release-notes/index.html
cuda download
https:/ /developer.nvidia.com/cuda-toolkit-archive
cudnn download
https://developer.nvidia.com/rdp/cudnn-download
nvcc-graphics driver difference
https://cloud.tencent.com/developer/article/1536738
Reference link: https://developer.aliyun.com/article/761648