Hey everyone! I am making a post today about who fits the bill to get the M1 MacBook for Data Science in 2020.
I am going to report my findings and then considering those factors describe who fits the bill to get one.
I’ve used the M1 MacBook Air for a few days and let’s get into what I like about the new Mac.
Macs are an absolute joy to develop code on. MacOS is built on top of UNIX. This makes lots of software engineering tasks really enjoyable. Also, for lots of machine learning tasks, it’s easier to do things like install packages or work in the terminal on a Mac/Linux compared to Windows in my opinion.
Programs for content creation (Premier, Photoshop) run very well too. Adobe products currently will need to run with Rosetta on the M1 as they don’t have the native support built out for Apple silicon yet. Even so, those programs run very fast. I can’t imagine it will get faster, but it will with the native support.
Overall, my experience is so much more pleasant than my windows desktop which has pretty extreme hardware for ML. Maybe this is personal preference, but I sacrifice some performance in the component cooling department and GPU department for the portability and user experience that this product delivers. For daily use, this will fly through anything that you throw at it.
Let’s get into the data science specific tasks that you will run on this: My experience with Jupyter notebook tasks has been incredible. For CPU-based tasks like data cleaning, preprocessing, extracting – this will eat Intel’s lunch, make them buy another lunch, and eat that one too. It’s that impressive.
Soon, I’ll be posting about the machine learning test comparison with previous MacBooks – though I do have a video discussing the results. In short, if you are a Mac fan, this is likely to be 2-3 times faster than your previous MacBook with those types of tasks. That was not even with the ARM-optimized version of Jupyter notebook which is supported with Miniconda – once I get that correctly installed (which admittedly isn’t too easy at this time) and do my demo on that, we are likely to see even further speed increases.
Alright, let’s get into what I’m not loving about these computers so far. The main thing is ARM-support. And specifically, it is trying to install tensorflow-macos which lots of people are still having issues with (myself included in that). One of the main reasons I bought it was to try out ARM architecture for ML and it has proven difficult so far. Saying that though, I do remain hopeful that these issues will get worked out as it matures – the product came out just a few weeks ago.
Another thing that I need to point out is the RAM limitation for data science tasks. I would recommend grabbing the 16GB version if you do plan on getting one of these so that you’ll be able to use it for a wider range of machine learning tasks. Make sure that 16GB is enough RAM for you – because it cannot be upgraded. Perhaps next year we will get a 32GB version and some of the ARM versions of software will be more developed – these may become the go-to laptops for data science.
So, who should get the M1 MacBook for Data Science?
People who are starting out or newer in data science and machine learning who plan to do lots of data cleaning and CPU-based machine learning, and perhaps training some shallow neural networks if you have the grit to get tensorflow-macos installed. It’s also for those who want to be using the laptop for lots of other things besides machine learning – coding, content creation you’ll find the user experience really breathtaking.
Those who are starting out in data science may not be jumping right into training deep neural networks. Training deep NN’s is the main use case for getting a good Nvidia GPU for which I will recommend a laptop or desktop with an Nvidia GPU at this time. If you aren’t doing that, then you’ll find the M1 a great option and again consider the 16GB version to expand the range of data science tasks and dataset sizes that will fly on the M1.
If you think the MacBook Pro or MacBook Air is right for you, you can get them on Amazon*:
MacBook Pro: https://amzn.to/35nvAA8
MacBook Air: https://amzn.to/35gqWE8
*Amazon Links are affiliate links that help generate revenue for the site.
Hope this helped clear up some questions about whether or not you should get the new M1 for your data science workloads. For those who don’t think the M1 is right for you at this time, I’ll be coming out with a more comprehensive laptop guide in a future video. Stay Tuned! Thanks for reading, and I’ll be back very soon with more machine learning related content. BYE!
If you enjoyed the article, consider sharing it with your friends/colleagues!