Mac Mini M1: Is 8GB Sufficient for Geospatial Data Science?
Written on
Introduction to Geospatial Data Science
When I began my journey into geospatial data science during my PhD in 2019, I quickly realized the immense computational power needed to handle large datasets. Anyone familiar with geospatial data understands that loading satellite images into Numpy arrays and merging them into complex data cubes can rapidly consume memory. To tackle this challenge, I invested in a high-performance Dell laptop equipped with 64GB RAM, a 6GB NVIDIA GPU, and a 9th generation Intel i7 processor, all for nearly 3,000 Euros. Despite its hefty weight of around 5kg, it proved to be a reliable partner.
My initial deep learning experiments with Fast.ai utilized the internal GPU, often running overnight. The noisy fan made it nearly impossible to sleep, and unfortunately, the battery deteriorated within two years. Over time, I transitioned most of the resource-intensive processing to the cloud, as the 6GB GPU could not keep up with newer architectures requiring 512x512 patches. Additionally, I began to leverage parallel processing capabilities on a high-performance cluster provided by the French National Centre for Space Studies (CNES).
After four years of intensive use, my Dell laptop, now running Windows 11, has started to lag despite its robust configuration. This leads me to my current topic. After completing my PhD in December and relocating to Brazil, I opted for a new "all-purpose" desktop. I chose a Mac Mini M1 but had to settle for the 8GB version due to the 16GB variant being out of stock. Although I had my doubts, I decided to experiment with this Apple Silicon, which cost me under 1,000 EUR.
Initial Impressions
Before diving into my experiments, I want to share my initial thoughts. Previously, I owned a 2011 iMac, which I sold once it stopped receiving updates. It served us well for nearly a decade with some DIY enhancements. However, the new M1 Mac truly excels in responsiveness. Tasks like browsing and writing are remarkably quick, a notable improvement over my Windows experience.
Moreover, the Mac Mini operates almost silently. I hardly notice it running, even with its internal cooling fan. Initially, I didn't plan to undertake heavy tasks, but I saw this as an opportunity for testing.
The Experiment
Thanks to Conda (specifically Miniconda), I was able to install all necessary libraries—including GDAL—compiled for the Arm64 architecture, eliminating the need for Rosetta code translation. I executed the waterdetect command while browsing the internet without interruptions. However, upon checking the activity monitor, I noticed the Mac Mini struggled with computations, peaking at 21GB of swap memory.
Ultimately, the Mac Mini completed the task in 17 minutes. In contrast, my Dell notebook accomplished the same task in just 4.5 minutes, showcasing almost a fourfold increase in performance compared to the Mac Mini.
Conclusion
In summary, the Mac Mini M1 with 8GB of RAM demonstrated its ability to perform a geospatial analysis task using the WaterDetect package, albeit with some challenges. The task peaked at 21GB of swap memory and took 17 minutes to finish. This suggests that while the Mac Mini M1 is functional for geospatial tasks, it may not be suitable for more demanding workloads requiring extensive memory.
That said, the responsiveness and quiet operation of the Mac Mini are impressive for everyday use, making it a solid choice for those seeking a compact, versatile desktop. However, for heavier tasks, considering a memory upgrade or opting for a more powerful machine would be wise.
Overall, I was pleasantly surprised by the Mac Mini's performance, especially given the substantial disparity in processing speed. I didn't expect it to handle the workload as well as it did, particularly when managing a 20GB swap file while still responding to commands.
Returning to the original question: Is a Mac Mini M1 with just 8GB adequate for geospatial data science? The answer is nuanced. If you are performing extensive in-house processing, a more powerful option, likely with 16GB of RAM, is advisable. However, for light prototyping and utilizing external cloud services for processing, the Mac Mini M1 can certainly suffice.
Chapter 2: Practical Comparisons
This video compares the M1 MacBook Air and Pro models, assessing their performance for data science tasks, including insights relevant for users considering the Mac Mini.
Chapter 3: Future Considerations
In this review, we explore the M1 Mac Mini in 2024, discussing whether it remains a worthwhile investment for potential buyers.