The world’s leading publication for data science, AI, and ML professionals.

M1 MacBook Pro vs. Intel i9 MacBook Pro – Ultimate Data Science Comparison

M1 vs. i9-9880H – performance comparisons with synthetic benchmarks, Python, Numpy, Pandas, and Scikit Learn.

Photo by Daniel Korpai on Unsplash
Photo by Daniel Korpai on Unsplash

M1 chip is revolutionary. Who could tell that a first generation chip from Apple will wipe the floor with decades worth of engineering from Intel? M1X and M2 chips make the future look promising, but there’s still a lot to enjoy from the current generation.

Today we’ll compare two machines in synthetic benchmarks, pure Python, Numpy, Pandas, and Scikit-Learn performance:

  • 16" MacBook Pro from 2019 – Intel Core i9–9880H, 16GB RAM, and AMD Radeon Pro 5500M (around $3K in the US)
  • 13" M1 MacBook Pro from 2020 – Apple M1 chip, 8GB of unified memory, and 8 GPU cores (around $1.3K in the US)

Keep in mind that these will be only simple programming and Data Science benchmarks, and a lot more could (and will) be done in every area of testing. Stay tuned for that.

Today’s article is structured as follows:

  • Synthetic Benchmarks – Geekbench and Cinebench
  • Comparing Pure Python Performance
  • Comparing Numpy Performance
  • Comparing Pandas Performance
  • Comparing Scikit-Learn Performance
  • Final Thoughts

Synthetic Benchmarks – Geekbench and Cinebench

Besides the performance differences you’ll see in a minute, there’s one important thing to note – 16" Intel-based MBP runs hot. Really hot. To the point that you can’t use it on your lap, at least not during summer. M1 Mac doesn’t have this issue. Its fan only kicked in in the last couple of minutes of the test.

Let’s start simple with a Geekbench score comparison in single core:

Image 1 - Geekbench comparison scores in single core (image by author)
Image 1 – Geekbench comparison scores in single core (image by author)

Just wow. The machine that costs almost three times as much doesn’t come anywhere near. Let’s see if the same applies to a multi core test. Keep in mind that Intel i9–9880H has 8 cores:

Image 2 - Geekbench comparison scores in multi core (image by author)
Image 2 – Geekbench comparison scores in multi core (image by author)

Ridiculous. Nothing more to add.

Let’s go over the GPU tests in Geekbench. This comparison isn’t 100% fair because the M1 Mac doesn’t have a dedicated GPU. Here are the results:

Image 3 - Geekbench comparisons in GPU (image by author)
Image 3 – Geekbench comparisons in GPU (image by author)

Almost double the score for the dedicated GPU, but that was expected. Next, let’s go over the comparison in Cinebench, both for single and multi core. Here’s for the single core:

Image 4 - Cinebench comparison scores in single core (image by author)
Image 4 – Cinebench comparison scores in single core (image by author)

Once again, the M1 chip takes the lead. Let’s see the results for a multi core test:

Image 5 - Cinebench comparison scores in multi core (image by author)
Image 5 – Cinebench comparison scores in multi core (image by author)

The machines are somewhat close, but the Intel-based Mac won this time. To conclude, both are very capable machines, but one would expect i9 to win every time with a significant difference, at least according to the price.

Winner – M1 MacBook Pro. It’s a better-performing machine in most cases, and it won’t melt your pants.


Comparing Pure Python Performance

Configuring the M1 chip for data science could be a pain in the bottom when doing it for the first time. The process isn’t the same as with the Intel chips, at least if you want to run everything natively. Luckily, here’s a step-by-step guide you can follow:

How to Easily Set Up M1 MacBooks for Data Science and Machine Learning

We’ll do a couple of relatively simple tasks in the pure Python test:

  • Create a list l containing 100,000,000 random integers between 100 and 999
  • Square every item in l
  • Take a square root of every item in l
  • Multiply corresponding squares and square roots
  • Divide corresponding squares and square roots
  • Perform an integer division of corresponding squares and square roots

As it’s a pure Python test, no third-party libraries are allowed. Here’s the code snippet:

And here are the results:

Image 6 - Pure Python performance comparison (image by author)
Image 6 – Pure Python performance comparison (image by author)

It isn’t such a significant difference, but still a clear victory for the M1 chip.

Winner – M1 MacBook Pro. Finished first for a third of the price.


Comparing Numpy Performance

Below you’ll find a list of tasks performed in this benchmark:

  • Matrix multiplication
  • Vector multiplication
  • Singular Value Decomposition
  • Cholesky Decomposition
  • Eigendecomposition

The original benchmark script was taken from Markus Beuckelmann on Github, and modified slightly, so both start and end time is captured. Here’s how the script looks like:

And here are the results:

Image 7 - Numpy performance comparison (image by author)
Image 7 – Numpy performance comparison (image by author)

Numpy is a different story. The test finished faster on the Intel chip, and the most likely reason is the Intel Maths Kernel Libraries (MKL) which isn’t found on the M1 chip.

Winner – Intel i9 MacBook Pro. Numpy just works faster. For now.


Comparing Pandas Performance

This benchmark is quite similar to the one done with pure Python. Identical operations were performed, but the results were combined to a single Pandas DataFrame:

  • Create an empty data frame
  • Assign it a column (X) of 100,000,000 random integers between 100 and 999
  • Square every item in X
  • Take a square root of every item in X
  • Multiply corresponding squares and square roots
  • Divide corresponding squares and square roots
  • Perform an integer division of corresponding squares and square roots

Here’s the code snippet:

Here are the results:

Image 8 - Pandas performance comparison (image by author)
Image 8 – Pandas performance comparison (image by author)

Intel i9–9880H is a very capable processor, but it looks like it can’t match the M1 chip in this task either.

Winner – M1 MacBook Pro. It’s both fast and silent.


Comparing Scikit-Learn Performance

Let’s stick to the basics here and do the following tasks:

  • Get the dataset from the web
  • Perform a train/test split
  • Declare a Decision tree model and find optimal hyperparameters (2400 combinations + 5-fold cross-validation)
  • Fit a model with optimal parameters

It’s a more or less standard model training procedure, disregarding testing out multiple algorithms, data preparation, and feature engineering.

Here’s the code snippet for the test:

And here are the results:

Image 9 - Scikit-Learn performance comparison (image by author)
Image 9 – Scikit-Learn performance comparison (image by author)

The results speak for themselves. I have nothing to add.

Winner – M1 MacBook Pro. Twice the speed for a third of the price.


Final Thoughts

The M1 chip is nothing short of revolutionary. I’ve done the benchmarks you’ve seen in the article, written and edited this entire piece, and streamed 40 minutes of Netflix – and still have 83% battery left!

M1 Macs don’t have a dedicated GPU, but that doesn’t have to be an issue. Chances are, you won’t use any laptop for resource and time consuming tasks, especially if we’re talking about deep learning. Doing it in the cloud is much more efficient.

So, for everyday data science and analysis work, M1 seems like a way to go – outstanding performance, stays cool, and the battery lasts up to two workdays. If you’re looking to buy a new Mac and everything you need is compatible with the M1 chip, I don’t see a reason to spend 2–4x more money and go with Intel.

What are your thoughts? How does your configuration compare to the M1 chip? Run the benchmarks and let me know.


Loved the article? Become a Medium member to continue learning without limits. I’ll receive a portion of your membership fee if you use the following link, with no extra cost to you.

Join Medium with my referral link – Dario Radečić


Learn More


Stay Connected


Related Articles