# For Nvidia Driver Installation Guides
###### tags: `Installation`

## Env.
Ubuntu 20.04 or 18.04
GPU Nvidia Tesla c2070

## Install Nvidia driver
In Ubuntu systems, drivers for NVIDIA Graphics Cards are already provided in the official repository. But since we are using the Tesla module graphics cards, we should dowload the driver from
[Nvidia driver](https://www.nvidia.com/download/index.aspx?lang=en-us). The latest supported driver version is ``390`` with cuda version ``9.1``.

### Install Dependencies
Software required for the runfile are officially listed [here](http://us.download.nvidia.com/XFree86/Linux-x86_64/384.69/README/minimumrequirements.html).

For Ubuntu we need:
1. **build-essential:** Include gcc, g++, cmake... and the other building tools.
2. **dkms:** For providing dkms support.

:::info
Since we don't need x-windows system, so we are not gonna install **xorg**.
:::

$ sudo apt install build-essential dkms

### Black Nouveau Driver
Nouveau is the open-soure Nvidia driver. In usual it is the default kernel module loaded in your system.

So edit the file at ``/etc/modprobe.d/blacklist.conf`` add the following contents:

blacklist nouveau
options nouveau modeset=0

then excute sudo ``update-initramfs -u`` and reboot the system.

#### Verify Configuration
To check if your changes were successful, you can use ``lspci -nnk`` and make sure the that the ``Kernel module in use:`` is not nouveau.

### Excuting the Runfile

$ wget http://us.download.nvidia.com/tesla/390.116/NVIDIA-Linux-x86_64-390.116.run
$ chmod +x NVIDIA-Linux-x86_64-390.116.run
$ sudo ./NVIDIA-Linux-x86_64-390.116.run

With the following steps to finish installation.
And reboot to apply change.

### Verify installation
Using nvidia-smi

kerwin@kerwin:~/cuda-driver$ nvidia-smi
Wed Aug  5 19:15:17 2020
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 390.116                Driver Version: 390.116                   |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  Tesla C2070         Off  | 00000000:01:00.0 Off |                  Off |
| 30%   49C   P12    N/A /  N/A |      0MiB /  6067MiB |      0%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|  No running processes found                                                 |
+-----------------------------------------------------------------------------+

## Install CUDA
:::warning
注意：CUDA 版本必須對應上Nvidia driver 的版本，像是這張卡最多就是上到390那對應的 cuda 版本就是9.1
:::

We can dowload the Linux version [here](https://developer.nvidia.com/cuda-91-download-archive?target_os=Linux&target_arch=x86_64&target_distro=Ubuntu&target_version=1604&target_type=runfilelocal)

$ sudo apt install gcc-6 g++-6

Dowload the four runfiles:

$ wget https://developer.nvidia.com/compute/cuda/9.1/Prod/local_installers/cuda_9.1.85_387.26_linux
$ wget https://developer.nvidia.com/compute/cuda/9.1/Prod/patches/1/cuda_9.1.85.1_linux
$ wget https://developer.nvidia.com/compute/cuda/9.1/Prod/patches/2/cuda_9.1.85.2_linux
$ wget https://developer.nvidia.com/compute/cuda/9.1/Prod/patches/3/cuda_9.1.85.3_linux

And excute these files in order

:::warning
當問你要不要安裝nvidia driver 的時候選擇 no ，因為我們已經裝過了。
:::

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-6 10
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-6 20

sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-6 10
sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-6 20

sudo update-alternatives --install /usr/bin/cc cc /usr/bin/gcc 30
sudo update-alternatives --set cc /usr/bin/gcc

sudo update-alternatives --install /usr/bin/c++ c++ /usr/bin/g++ 30
sudo update-alternatives --set c++ /usr/bin/g++

Add the following contain to ~/.bashrc

PATH="/usr/local/cuda-9.1/bin:$HOME/bin:$HOME/.local/bin:$PATH"
LD_LIBRARY_PATH=/usr/local/cuda-9.1/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}