Hello Hilal Abbood Al-Libawy I will give you some links that will help you answer your question. Have a look hopefully you will find something that will create a spark.

https://docs.nvidia.com/deeplearning/performance/dl-performance-getting-started/index.html#:~:text=1.-,Overview,to%20efficiently%20use%20GPU%20resources.

https://timdettmers.com/2018/12/16/deep-learning-hardware-guide/

https://towardsdatascience.com/what-is-a-gpu-and-do-you-need-one-in-deep-learning-718b9597aa0d

let us know if these were helpful

All the best!

Hello Hilal Abbood Al-Libawy I will give you some links that will help you answer your question. Have a look hopefully you will find something that will create a spark.

https://docs.nvidia.com/deeplearning/performance/dl-performance-getting-started/index.html#:~:text=1.-,Overview,to%20efficiently%20use%20GPU%20resources.

https://timdettmers.com/2018/12/16/deep-learning-hardware-guide/

https://towardsdatascience.com/what-is-a-gpu-and-do-you-need-one-in-deep-learning-718b9597aa0d

let us know if these were helpful

All the best!

Srinivas Talasila I will say all are useful.

https://www.researchgate.net/deref/https%3A%2F%2Ftowardsdatascience.com%2Fwhat-is-a-gpu-and-do-you-need-one-in-deep-learning-718b9597aa0d

https://www.researchgate.net/deref/https%3A%2F%2Ftimdettmers.com%2F2018%2F12%2F16%2Fdeep-learning-hardware-guide%2F

The above links will give you a overview literature on GPU and why we need one in deep learning.

The below link is from NVIDIA (computer gaming company that invented GPU) gives their technical input on GPU and deep learning.

https://www.researchgate.net/deref/https%3A%2F%2Fdocs.nvidia.com%2Fdeeplearning%2Fperformance%2Fdl-performance-getting-started%2Findex.html%23%3A~%3Atext%3D1.-%2COverview%2Cto%2520efficiently%2520use%2520GPU%2520resources

I hope you find this helpful

Regards

I said those links are very useful...

For experience, I would discard a laptop with a discrete GPU. For deep learning, you need the GPU to run almost at full load for hours. A laptop usually does not have the necessary cooling to actually take advantage of the real performance of most GPUs. The few that can actually cool them are heavy, bulky, and expensive. On top of that, if the laptop runs hot for long periods of time, components like the batteries are going to degrade much faster.

My suggestion is to buy a desktop PC with a dedicated graphics card and a cheap laptop to access the desktop PC remotely.

The parameters to look in a GPU are:

-the brand: only Nvidia cards take advantage of the CUDA acceleration for deep learning

-VRAM: if you want to train models with high-resolution images or that uses multiple frames (recurrent networks) you should invest in a GPU with a large VRAM. However, you do not need the most expensive cards with large VRAM necessarily, there are many techniques to save GPU memory.

-Tensor Cores: More there are in the card, the faster it will be.

All the best!

I use GPU to train the CNN models

However, I have seen people use FPGA Accelerator

check this paper

Chapter Boosting Convolutional Neural Networks Performance Based on ...