Step By Step Installing TensorFlow 2 on Windows 10 ( GPU Support, CUDA , cuDNN, NVIDIA, Anaconda)

It's easy if you fix your versions compatibility

System:

Windows-10
NVIDIA Quadro P1000

Step1:
Download & Install ( I used 2019 Version Latest at the time of writing)
https://visualstudio.microsoft.com/vs/community/

Step2:
Get the NVIDIA right driver and install
(For me it is NVIDIA QUADRO P1000)
https://www.nvidia.com/download/index.aspx?lang=en-us

Step3:
Download and Install Cuda toolkit
I downloaded 11.1 version
https://developer.nvidia.com/cuda-toolkit-archive

Step5:
Download and Install cuDNN
( Version compatibility must, I used 8.0.4 for 11.1 Cuda toolkit)
https://developer.nvidia.com/cudnn

Step6:
unzip cuDNN folder
go to cuda/bin directory
copy all files
go to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin
pate copied files here
go to cuda/include directory
copy all files and paste it inside
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\include directory
go to cuda/lib and copy all files
paste copied files inside
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\lib

Step7: Important alert
Go to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin
rename cusolver64_11.dll to cusolver64_10.dll
Tensorflow will look for cusolver64_10.dll and can not find it if you do not rename it.

Step8:
Open windows command prompt
Set environment variable path
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\bin;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\extras\CUPTI\lib64;%PATH%
SET PATH=C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v11.1\include;%PATH%

Step9:
Download and Install anaconda
https://www.anaconda.com/products/individual

Step10:
open anaconda command propmt
create a virtual environment and activate
conda create — name tfgpu python=3.8.0
conda activate tfgpu
pip install tensorflow-gpu
This command installed tensorflow_gpu-2.4.0 version for me

Step11:
Check if tensorflow installed with gpu
import tensorflow as tf
Command below will print which gpu it is using
tf.test.gpu_device_name()

Check https://www.tensorflow.org/install/gpu if version compatibility issue