ubuntu setup nvidia development environment on a single machine(not virtual machine)

最新推荐文章于 2025-04-11 15:09:08 发布

WhyNotFocus

最新推荐文章于 2025-04-11 15:09:08 发布

阅读量292

点赞数

分类专栏：软件安装文章标签： ubuntu

本文链接：https://blog.youkuaiyun.com/mos1896/article/details/120667877

版权

软件安装专栏收录该内容

6 篇文章

订阅专栏

本文档详细记录了在Ubuntu18.04系统上，使用NVIDIA驱动460，CUDA10.2及CuDNN8.0.3.33的安装步骤。首先通过软件与更新来安装NVIDIA驱动，并重启电脑以激活。接着安装CUDA10.2，然后安装与CUDA版本匹配的CuDNN。在完成所有设置后，验证了GPU测试项目，发现系统中已有的GCC和G++版本为7，而之前的经验可能需要4.8版本。不过，当前配置并未引起问题。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

ubuntu 18.04 ,gtx1080ti
setup nvidia driver using software&update building in ubuntu,refer to address1,i choose nvidia-driver-460,notice that we need to restart the computer to make it work.
setup cuda10.2,refer to address 2
setup cudnn:libcudnn8_8.0.3.33-1+cuda10.2,refer to address 2
when i prepare the anaconda(which i copy from my old directory),pycharm environment,i run the GPUtest project in /home/myname/SetupSoftware,it worked,notice that my gcc and g++ is version 7,however my old memory tells me that we should use gcc and g++ version 4.8,maybe it is due to the different cuda and cudnn version.