新的 Amazon EC2 集群 GPU 实例体验如何?
亚马逊发布了 EC2 - 集群 GPU 实例,我想知道您对此有何体验?它是否稳定,在部署 CUDA 代码之前是否需要大量时间来安装新的驱动程序、SDK 等?
Amazon released EC2 - Cluster GPU Instances and I wonder what's your experience with it? Is it stable, does it require a lot of time to install new drivers, SDK etc. before you deployed your CUDA code?
如果你对这篇内容有疑问,欢迎到本站社区发帖提问 参与讨论,获取更多帮助,或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱,如果其他用户或者作者回复了您的评论,将不能在第一时间通知您!
发布评论
评论(1)
我尚未部署 GPU 实例,但我可以告诉您,操作系统映像已为您设置了驱动程序。
现在,关于安装 CUDA 以及让代码运行,那是另一回事了。如果您根本没有尝试过 EC2 那么我可以在普通实例上告诉您 - 我可以安装 gcc/g++ 和 svn;设置存储库并在 5-10 分钟内运行我的代码。
编辑:我正在查看文档,发现了这个: http://docs .amazonwebservices.com/AWSEC2/latest/UserGuide/Cluster_GPUs_Install_Driver.html#d0e18924 这讨论了重新安装或更新 NVIDIA 驱动程序
I haven't yet deployed a gpu instance, but I can tell you that the OS image already has the drivers setup for you.
Now in terms of installing CUDA, and getting your code ran thats anotother stody. If you haven't tried EC2 at all then I can tell you on a normal instance - I can install gcc/g++ and svn; setup a repository and have my code run in 5-10 minutes.
EDIT: I was looking through the documentation and found this: http://docs.amazonwebservices.com/AWSEC2/latest/UserGuide/Cluster_GPUs_Install_Driver.html#d0e18924 this talks about reinstalling or updating the NVIDIA drivers