Ubuntu16.04上升级NVIDIA显卡驱动及安装CUDA10.0操作步骤

今天药忘吃喽~ 2022-12-22 11:23 177阅读 0赞

Ubuntu 16.04上已装有CUDA 8.0，现在想再安装CUDA 10.0,由于已安装的显卡驱动版本396.54不支持CUDA 10.0，因此安装CUDA 10.0之前需要先升级显卡驱动到410及以上版本。

可在[https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html][https_docs.nvidia.com_cuda_cuda-toolkit-release-notes_index.html]中查看对应的显卡驱动支持的CUDA版本。

1. 升级显卡驱动版本到430，依次执行如下命令后，重启电脑，

sudo apt-get update
    sudo add-apt-repository ppa:graphics-drivers/ppa
    sudo apt install nvidia-430

通过终端nvidia-smi命令查看，可查看显卡驱动信息，结果如下图所示，升级成功:

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70][]

2. 安装CUDA 10.0：

(1).从https://developer.nvidia.com/cuda-10.0-download-archive?target\_os=Linux&target\_arch=x86\_64&target\_distro=Ubuntu&target\_version=1604 下载runfile(local)的CUDA，即cuda\_10.0.130\_410.48\_linux.run;

(2).执行如下命令，安装CUDA 10.0:

sudo sh cuda_10.0.130_410.48_linux.run

在终端所做的选择：

Do you accept the previously read EULA? accept
    Install NVIDIA Accelerated Graphics Driver for Linux-x86_64 410.48? n
    Install the CUDA 10.0 Toolkit? y
    Enter Toolkit Location [ default is /usr/local/cuda-10.0 ]: Enter
    Do you want to install a symbolic link at /usr/local/cuda? y
    Install the CUDA 10.0 Samples? y
    Enter CUDA Samples Location [default is /home/xxxx ]: Enter

(3).验证CUDA 10.0是否安装正确,依次执行如下命令：

cd /usr/local/cuda/samples
    sudo make -k
    ./1_Utilities/deviceQuery/deviceQuery

执行结果如下图所示，说明CUDA 10.0安装正确：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70 1][]

3. 安装cuDNN:

(1).从https://developer.nvidia.com/rdp/cudnn-archive下载：cuDNN v7.3.0 Runtime Library for Ubuntu16.04 (Deb), cuDNN v7.3.0 Developer Library for Ubuntu16.04 (Deb), cuDNN v7.3.0 Code Samples and User Guide for Ubuntu16.04 (Deb)即libcudnn7\_7.3.0.29-1+cuda10.0\_amd64.deb、libcudnn7-dev\_7.3.0.29-1+cuda10.0\_amd64.deb、libcudnn7-doc\_7.3.0.29-1+cuda10.0\_amd64.deb；

(2).安装.deb包：默认cudnn动态库会安装在/usr/lib/x86\_64-linux-gnu目录下，cudnn.h会安装到/usr/include目录下，cudnn\_samples\_v7会安装到/usr/src目录下

sudo dpkg -i libcudnn7_7.3.0.29-1+cuda10.0_amd64.deb
    sudo dpkg -i libcudnn7-dev_7.3.0.29-1+cuda10.0_amd64.deb 
    sudo dpkg -i libcudnn7-doc_7.3.0.29-1+cuda10.0_amd64.deb

4. Export CUDA路径，打开/etc/profile文件：

(1).在原有LD\_LIBRARY\_PATH值的基础上增加/usr/local/cuda-10.0/lib64；

(2).在原有PATH值的基础上增加/usr/local/cuda-10.0/bin；

5.验证cuDNN是否安装正确：

(1).进入到/usr/src/cudnn\_samples\_v7/mnistCUDNN目录下；

(2).依次执行如下命令：

sudo make
    ./mnistCUDNN

执行结果如下图所示，说明安装正确：

![watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70 2][]

**GitHub**：[https://github.com/fengbingchun/CUDA\_Test][https_github.com_fengbingchun_CUDA_Test]

[https_docs.nvidia.com_cuda_cuda-toolkit-release-notes_index.html]: https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70]: /images/20221120/2a9c55e6f66f4eb081e567768ee1527f.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70 1]: /images/20221120/fc45d121621849b2b2ddf01110cab1bd.png
[watermark_type_ZmFuZ3poZW5naGVpdGk_shadow_10_text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L2ZlbmdiaW5nY2h1bg_size_16_color_FFFFFF_t_70 2]: /images/20221120/e53a0d7499ff4e46adc916fc480f97cc.png
[https_github.com_fengbingchun_CUDA_Test]: https://github.com/fengbingchun/CUDA_Test