当前位置: 首页 > 工具软件 > Titon Toolkit > 使用案例 >

Ubuntu16.04+cuda8.0+GTX TITAN X安装配置

左恺
2023-12-01

安装环境

显卡型号:Nvidia GeForce GTX TITAN X(pascal)

系统:ubuntu16.04

 

1、给出cuda下载地址(含历史版本):

https://developer.nvidia.com/cuda-toolkit-archive

选择合适的版本。

 

2、下载cuda(使用run类型),依次选择:

Operating Systemlinux

Architecturex86_64

Distributionubuntu

Version16.04

Installer Typerunfile(local)

 

3、安装显卡驱动:

sudo apt-get install nvidia-367

 

该方法不需要关闭nouveaulightdm,如果出现依赖问题可以尝试换源。

在楼主的环境中使用阿里云的源安装成功。

安装完成后,输入nvidia-smi测试。结果如下:

 

ubuntu@ubuntu-pc:~$ nvidia-smi

Fri Jan 19 22:02:51 2018      

+-----------------------------------------------------------------------------+

| NVIDIA-SMI 384.111                Driver Version: 384.111                   |

|-------------------------------+----------------------+----------------------+

| GPU  Name       Persistence-M| Bus-Id       Disp.A | Volatile Uncorr. ECC |

| Fan  Temp Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |

|===============================+======================+======================|

|  0  TITAN X (Pascal)    Off | 00000000:02:00.0  On |                  N/A |

| 23%   26C   P0    53W / 250W |    182MiB / 12188MiB |      0%     Default |

+-------------------------------+----------------------+----------------------+

                                                                              

+-----------------------------------------------------------------------------+

| Processes:                                                      GPU Memory |

| GPU       PID   Type  Process name                            Usage      |

|=============================================================================|

|   0      1081      G  /usr/lib/xorg/Xorg                           138MiB |

|   0      1860      G  compiz                                        41MiB |

+-----------------------------------------------------------------------------+

 

3、安装cuda

Run `sudo shcuda_8.0.44_linux.run`   and   Follow the command-line prompts

运行cuda文件,根据提示输入一些yn,主要是一些默认安装位置的选择和相关库的安装。因为前面已经安装过显卡驱动,该步不需要再安装显卡驱动,选择n即可。

 

4、测试cuda

打开cuda8.0 Samples默认安装路径(在第三步的相关选项中会提示设置,默认为:/home/ubuntu/NVIDIA_CUDA-8.0_Samples,ubuntu为用户名)编译:

sudo make all -j4

 

某些教程可能提到gcc版本过高的问题,我的ubuntu16.04默认gcc版本为5.4,编译没有出现问题。

 

编译完成后,在当前目录继续输入:

cd bin/x86_64/linux/release

./deviceQuery

出现下面报告,说明安装成功。

 

ubuntu@ubuntu-pc:~/NVIDIA_CUDA-8.0_Samples/bin/x86_64/linux/release$./deviceQuery

./deviceQuery Starting...

 

 CUDA Device Query (Runtime API) version(CUDART static linking)

 

Detected 1 CUDA Capable device(s)

 

Device 0: "TITAN X(Pascal)"

 CUDA Driver Version / Runtime Version          9.0 / 8.0

 CUDA Capability Major/Minor version number:    6.1

 Total amount of global memory:                 12189 MBytes (12780699648bytes)

 (28) Multiprocessors, (128) CUDA Cores/MP:     3584 CUDA Cores

 GPU Max Clock rate:                            1531 MHz (1.53 GHz)

 Memory Clock rate:                             5005 Mhz

 Memory Bus Width:                              384-bit

 L2 Cache Size:                                 3145728 bytes

 Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536),3D=(16384, 16384, 16384)

 Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers

 Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers

 Total amount of constant memory:               65536 bytes

 Total amount of shared memory per block:       49152 bytes

 Total number of registers available per block: 65536

 Warp size:                                     32

 Maximum number of threads per multiprocessor:  2048

 Maximum number of threads per block:           1024

 Max dimension size of a thread block (x,y,z): (1024, 1024, 64)

 Max dimension size of a grid size   (x,y,z): (2147483647, 65535, 65535)

 Maximum memory pitch:                          2147483647 bytes

 Texture alignment:                             512 bytes

 Concurrent copy and kernel execution:          Yes with 2 copy engine(s)

 Run time limit on kernels:                     Yes

 Integrated GPU sharing Host Memory:            No

 Support host page-locked memory mapping:       Yes

 Alignment requirement for Surfaces:            Yes

 Device has ECC support:                        Disabled

 Device supports Unified Addressing (UVA):      Yes

 Device PCI Domain ID / Bus ID / location ID:   0 / 2 / 0

 Compute Mode:

    < Default (multiple host threads can use ::cudaSetDevice() withdevice simultaneously) >

 

deviceQuery, CUDA Driver = CUDART,CUDA Driver Version = 9.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 =TITAN X (Pascal)

Result = PASS

 

5、添加环境变量

使用gedit编辑bashrc文件:

sudo gedit ~/.bashrc

在最后添加两行(cuda安装在默认位置时):

exportPATH=/usr/local/cuda-8.0/bin:$PATH

exportLD_LIBRARY_PATH=/usr/local/cuda-8.0/lib64:$LD_LIBRARY_PATH

然后使用:source ~/.bashrc,使修改立即生效。

 

接下来可以继续安装cudnn和其他库。

 

 类似资料: