Disclaimer: I wrote this post several months ago and did not publish it for unknown reason, I assume that the information below is still relevant and correct.
While trying to install the machine, I started my research by reading Traun Leyden's great blog post: http://tleyden.github.io/blog/2014/10/25/cuda-6-dot-5-on-aws-gpu-instance-running-ubuntu-14-dot-04/ (you should too)
Amazon offers two types of machines that includes GPUs (https://aws.amazon.com/ec2/instance-types/#g2)
This will be quick and simple:
1. Make sure you are fully up-to-date
2. Reboot the machine (so it will load the new kernel)
3. Install the kernel's header files
5. Update the package manager and install CUDA
7. You can also compile and run a sample to make sure everything is ready:
That's all!
- Tal Kain
While trying to install the machine, I started my research by reading Traun Leyden's great blog post: http://tleyden.github.io/blog/2014/10/25/cuda-6-dot-5-on-aws-gpu-instance-running-ubuntu-14-dot-04/ (you should too)
Amazon offers two types of machines that includes GPUs (https://aws.amazon.com/ec2/instance-types/#g2)
High-performance NVIDIA GPUs, each with 1,536 CUDA cores and 4GB of video memoryWhile writing this post, I used the g2.2xlarge machine, but you can also use the 8xlarge.
This will be quick and simple:
1. Make sure you are fully up-to-date
sudo apt-get update && sudo apt-get upgrade && sudo apt-get dist-upgradeWhen prompted, choose the "install maintainer package...."
2. Reboot the machine (so it will load the new kernel)
3. Install the kernel's header files
sudo apt-get install -y linux-image-extra-virtual linux-headers-`uname -r`3. Configure the xorg-edgers PPA
sudo add-apt-repository ppa:xorg-edgers/ppa4. Install the NVIDIA's CUDA repository package
sudo apt-get update
OUTPUT_PATH=/tmp/cuda.deb
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_7.0-28_amd64.deb -O ${OUTPUT_PATH}
dpkg -i ${OUTPUT_PATH}
5. Update the package manager and install CUDA
sudo apt-get update
sudo apt-get install cuda
6. Make sure everything is installed as expected:
sudo nvidia-smi -a
7. You can also compile and run a sample to make sure everything is ready:
cd /usr/local/cuda/samples/1_Utilities/deviceQuery
make
./deviceQuery
That's all!
- Tal Kain
Well, pretty well... thanks for the detailed explanation, for NVIDIA GK104 everything works fine, the only one for NVidia drivers download for Windows http://bitdrivers.com/manufacturers/nvidia (64Х) since the previous ones are outdated and weren't connecting with my OS/
ReplyDelete