Welcome back smartpants!
CUDA Cores are parallel processors. A CPU may have 2 or 8 or dozens of cores, but a nVidia GPUs have CUDA cores ranging from hundreds to thousands.
Lets start putting these minions to use!
In this post, we will:
- Find out how many CUDA Cores you have.
- Run some programs that utilize CUDA Cores.
By the end of this tutorial, you will have the skills to use the CUDA cores in your GPU.
Installation
Install a cross platform tool called Anaconda Navigator. Installation of Anaconda and its usage is not covered in this tutorial.
Follow these steps:
- Install Anaconda
- Install numba if you don’t already have it using
conda install numba - Install the CUDA toolkit version 8.0 or higher
conda install cudatoolkit - Open Spyder or Jupyter Notebook or VS Code. Whatever tickles your fancy.
We are set! Now lets check how many CUDA Cores we have.
How many of CUDA Cores do I have?
In Ubuntu 20.04, I was able to find the number of CUDA cores like this:
myuser@pc-ThinkPad-P50:~$ nvidia-settings -q CUDACores -t 640 640
But this is not enough because my Windows users may need a to check their CUDA cores as well. They can install GPUZ or CPU-Z to check that.
Confirm that we have a CUDA Supporting GPU(s)
Lets check if my Lenovo P50 has even got CUDA support

Ok, then which GPI do I have and is it supported by Numba?

This is good. We are covered by CUDA Cores. This means that you can now continue and start working on using those cuda cores of yours 😉
Thanks for reading.