What’s up everyone! This post is going to be for those in the market for a laptop, and more specifically those who want to use that laptop for machine learning.
As I always say, if you want to do intense deep learning, you probably are going to want a desktop computer that will be able to have ample air cooling so that you can train models for extended periods of time. Also overclocking your components becomes a lot simpler with a desktop and enhanced cooling. I have a ton of videos on the channel if you think that describes you. So desktops might have an advantage in a pure performance battle, but laptops offer a huge advantage for many us: they’re mobile. In order to be mobile, we are willing to sacrifice some performance. You might be going to college to study this material or want to work on analytics while you’re watching TaskMaster – doesn’t matter.
The intensity that you want to use it for machine learning might vary – some might use it for a single class, whereas others want to be training models every night.
What things should you look for in this decision as applied to machine learning? Well, you are choosing a laptop to be portable so you don’t want one of the laptops with a full-fledged desktop card – that will hinder your mobility. They make laptops that have the same specs as my deep learning workstation. I’m talking about the Alienware 17 R5 – this thing is absurd: RTX 2080, 32GB Ram, and an Intel Core i9 processor. If you are interested and have a high budget, you should check it out on Amazon: https://amzn.to/38kJns8
It will give you extremely high performance, but also, it weighs 15 pounds. And it’s by no means cheap. Some of these types of hardware-heavy laptops even require two power brick connections. My thinking is that we are going for a laptop due to the portability – if you want something like this you might as well buy a desktop.
So, we want killer specs with a high portability factor. Specifically, what components should we home in on.
First up I would say RAM is really important with laptops. Lots of them try to sell you on 8GB of RAM and unless that’s a Mac M1 where that’s a highly unified 8GB of RAM integrated into the rest of the components very tightly you will likely find 8gb to be very limited. You will ideally want 16GB of RAM for your laptop. This is just going to expand the range of tasks that you will be able to complete with the computer.
Next up is graphics cards. Graphics cards are going to give you that really nice acceleration with the deep learning platform, CUDA, if you have an Nvidia chip. Nvidia is where it’s at with deep learning. If you are wondering about the new M1 MacBook and how that fares against an Nvidia chip, check out my video where I compare TensorFlow on my M1 with my deep learning desktop. Spoiler alert: while the M1 put a good fight, the Nvidia definitely ate its lunch. So Nvidia chips are still the way to go for deep learning hardware.
Nvidia makes notebook versions of their graphics cards so while you might have a GTX 1070 in a desktop, you will have the mobile version in a laptop oftentimes. These are a little more compact and thus slightly less performant than the full desktop version. However, they still do pretty well.
I should also mention that cost is highly correlated with performance of the component, so this advice is geared to normal people who want to become involved in the field of study who aren’t spending thousands and thousands of dollars to get the absolute best deep learning laptop.
So lots of factors here to think about: high amount of RAM, a respectable NVidia GPU, CPU, and a great portability factor since these are laptops after all.
I look at a lot of different laptops since people ask me what they should get for their specific application. All things considered for machine learning my top overall pick for deep learning in 2020 is the ROG Zephyrus M (check it out on Amazon: https://amzn.to/3nyorEt)
It has tons of RAM with space to add even more, it has an outstanding GPU, and a great processor. This is an Nvidia card, so you’ll gain experience with the CUDA libraries and get all that GPU acceleration. If you end up building a desktop workstation down the road, that knowledge will definitely come in handy.
Thanks so much for reading, have a good day. BYE!
Disclosure: AI Buzz Media is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to amazon.com