chhanganivarun/AMD NVIDIA CUDA.txt

## AMD NVIDIA CUDA.txt
Steps for installing Ubuntu 20.04 on an AMD Ryzen with a cuda compatible nvidia device
1. Install Ubuntu without any third party apps. The installation should happen without any problem.
2) Restart ur system as is suggested after installation. When the grub menu appears, press 'e' over Ubuntu and append nomodeset to the line containing quick splash ( Only then will you be able to render ur display as you don't have an nvidia-driver installed yet.)
3) Then, run ubuntu-drivers devices to check if a new nvidia-driver is available. I installed driver 450 (is compatible with cuda 10.1)
4) If such a driver does not show up, sudo apt update. The driver should then show up in ubuntu-device drivers.
5) Install using sudo apt install nvidia-driver-450
6) Run nvidia-smi to check if it was properly installed.
7) Restart once.
8) For CUDA installation, go to tensorflow docs. Then have documented it properly. Follow all the steps for CUDA Installation there(1st step for installing nvidia driver is not reqd as you have already installed it).
RUN ALL COMMANDS FOR CUDA INSTALLATION EXCEPT THE LAST ONE. It installs an nvidia-driver which it finds suitable and that would definitely cause errors. Instead install cuda-toolkit. I will attach the link to do that.
9) After that install Cudnn an follow the steps in the nvidia documentation for having it linked with Cuda.
10) Install tensorflow-gpu==2.3.0
	Steps for installing Ubuntu 20.04 on an AMD Ryzen with a cuda compatible nvidia device
	1. Install Ubuntu without any third party apps. The installation should happen without any problem.
	2) Restart ur system as is suggested after installation. When the grub menu appears, press 'e' over Ubuntu and append nomodeset to the line containing quick splash ( Only then will you be able to render ur display as you don't have an nvidia-driver installed yet.)
	3) Then, run ubuntu-drivers devices to check if a new nvidia-driver is available. I installed driver 450 (is compatible with cuda 10.1)
	4) If such a driver does not show up, sudo apt update. The driver should then show up in ubuntu-device drivers.
	5) Install using sudo apt install nvidia-driver-450
	6) Run nvidia-smi to check if it was properly installed.
	7) Restart once.
	8) For CUDA installation, go to tensorflow docs. Then have documented it properly. Follow all the steps for CUDA Installation there(1st step for installing nvidia driver is not reqd as you have already installed it).
	RUN ALL COMMANDS FOR CUDA INSTALLATION EXCEPT THE LAST ONE. It installs an nvidia-driver which it finds suitable and that would definitely cause errors. Instead install cuda-toolkit. I will attach the link to do that.
	9) After that install Cudnn an follow the steps in the nvidia documentation for having it linked with Cuda.
	10) Install tensorflow-gpu==2.3.0