Best laptop for Deep Learning and AI

Today one of my junior asked me that he is going to purchase a laptop to practice deep learning and artificial intelligence. I suggest some points, which I thought you should also know before buying laptops for deep learning, artificial intelligence, machine learning or even general data science practice.

I had a call for around 5-10 minutes to explain those points, that he needs to check before buying laptops for deep learning or artificial intelligence. Let me break that entire conversation into some headings for your better understanding.

Graphics Card or GPU

Personally, I am using a laptop with 4GB Nvidia 1050ti graphics card GPU. I bought this laptop around 2018. At that time (2018), that much graphics card (4GB) was more than enough to practice Deep learning, machine learning, and even artificial intelligence.

I could run or fine-tune almost any deep-learning model with that laptop. But in 2023 artificial intelligence field is evolving rapidly. Nowadays size of a good deep-learning model is with minimum of 100 million parameters.

So to load this much big AI model and finetune for let’s say at least 10,000 data, your GPU of 4gb ram will raise its hand. Standing at 2023, I feel that a best laptop to practice deep learning should have at least 15 GB of GPU. That is in 2023. Maybe in 2026 15 GB of GPU will not be sufficient to run deep learning models even for batch size 2.

This is the reason, I don’t think you should give much priority to GPUs for your future deep-learning laptop. You can use free Google Collaboratory or buy their subscription to get better runtime.

Still if you determined to buy GPU based laptop. Then before buying a gaming laptop especially for deep learning, you must check supported NVIDIA graphics card for Cuda installation. Else you end up spending lots of money but your purpose will not be solved. You can check Cuda supported graphics card list here at this link.

Processor or CPU

If your GPU doesn’t support any model and if you don’t want to run it in Google colabotory or AWS ( Amazon Web Services), then the last option remain CPU.

Along with that lots of old machine learning models runs only on CPU. Only tensorflow and torch these are the two framework support GPU processing. A good processor can also help you to do multi processing. So if you are looking for best laptop for Machine learning then you must use latest processor.

The time when I bought my laptop to practice AI, on that time Intel i7 processor was the latest Intel processor. So I went with that processor. Trust me that i7 processor performing extreamly well till today, even with heavy deep learning models like: Tecotron2, MobileNet, BERT, T5, etc.

RAM

When you are planning to buy a laptop for data science specially for artifical intelligence, don’t cut your budget in this segment.

If your model does not support ( if you are not using tensorflow or torch based model) GPU or your GPU does not support ( less GPU memory) the model, you need to run models in processor.

Now while running models in CPU, your RAM memory will be utilized. Your input data and model will first load into RAM then CPU will process those from RAM. This is the reason you need plenty size of RAM.

Generally a gaming laptop comes with 8gb of RAM, but you can expand that ram upto 32 or even more GBs. In my case, my laptop came with 8gbs of ram. Additionally I purchased one 16 GB Ram and inserted in the second slot. So in total my system has ram size of 24 GB now.

Now question is 24 GB of ram is enough for deep learning and artificial intelligence or not. The answer is till now I am running almost all models flawlessly. I don’t know about the future.

One suggestion: if you are facing issue with ram size, then try to reduce your training data and batch size. This trick might work for you.

Ports

If you are determined to purchase a laptop for practicing deeplearning, then you should know about the Display port or in short DP port.

As I told earlier, there is no limit of GPU. If some GPU is working fine today, may be the same GPU will be obsolute tomorrow. This is happening because day by day model size is increasing. For example today you purchased a gaming laptop with 4gb of GPU memory, and tomorrow you find that 4gb is not enough, you need 8 GB of graphics card.

You can not throw your laptop and buy another right. In this case dp port will help you to connect with external GPUs.

So what is dp port? Dp port is one type of port like usb or HDMI. This dp port will help you to connect external GPUs. This way you can extend your GPU memory and your laptop will not outdated soon.

Unfortunately, my laptop does not have dp port. I was not aware of this befor. But you should know this before buying laptop for deep learning.

Hardware

If you are buying a gaming laptop for machine learning and ai, obviously you will give some load to it. Now if you give load to the laptop, it will definitely become hot. So you need to check its cooling system before buying any laptop to run deep learning models.

You must buy a laptop with atleast dual cooling fan with with 4 cooling solution. 4 cooling solution means there are 4 window from where hot air will be passed. These 2 cooling fan and 4 cooling windows will keep your laptop cool even when you run any deep learning model for hours.

I just filtered laptops on amazon.com with an Intel i9 processor. I found the following list of laptops. If you want to stay within your budget, you can also consider laptops with an i7 processor. With i7 you can expect less price.

My Suggestion

So which laptop is best for ML and AI. I will still say if you are looking for best laptop for deep learning practice, you should buy a laptop without GPU. Motherboard inbuilt graphics card can be there but not dedicated GPU like GeForce, Tesla etc.

Why I am saying this because if you are buying a laptop with GPU, your maintenance cost will exponentially increase with your laptop hardware. Also, gaming laptops are heavy in weight, so there can be portability problem.

You can simply buy a good cpu laptop and extend your ram size. With this configuration you can test your deep learning or machine learning code in your local laptop with low batch size and 1 or 2 epochs. Once your are ready with your code, you can run the same code in Google colab with larger data, higher batch size and epochs. Once you successfully trained your deep learning model, download it and use it in your local laptop.

Another minor but important thing is that for some laptop you will not be able to replace your keyboard because of its unavailability after 2 years. Better you also buy a replacement keyboard along with new laptop . It is obvious that keyboard of your laptop will stop working after 2 years or even before that.

This is the reason you should understand what is required for you, if you are planning to use it in your home only, then instead of laptops, you must buy a desktop for deep learning.

So this is it. In this article, I shared some points which you need to consider before choosing a Best laptop for artificial intelligence and data science. If you have any questions or suggestions regarding this post, please let me know in the comment section below.