Member-only story
Is a Mac Mini M1 With Just 8Gb Enough for Geospatial Data Science?
Assessing the true power of Apple Silicon architecture for geospatial analysis.
Introduction
When I started working with geospatial data science during my PhD in 2019, I knew that the volume of data would require significant computational power. Those who work with geospatial data know that it’s easy to fill up memory by loading satellite imagery into Numpy arrays and combining them into data cubes with multiple dimensions. To handle this, I purchased a high-spec notebook from Dell, with plenty of memory (64Gb), a 6Gb NVIDIA GPU, and a 9th generation Intel i7 processor. The cost was almost 3,000 Euros, but I believed it was necessary for the journey ahead. Despite its weight of almost 5kg, it served as a good companion.
My first deep learning experiments with Fast.ai were performed using the internal GPU, running all night long. The fan worked at full throttle, making it almost impossible to sleep due to the loud noise in the room. This also resulted in the battery being ruined in just 2 years. As time passed, I moved most of the heavy processing to the cloud, as the 6Gb GPU was not sufficient for newer architectures with 512x512 patches. Conventional processing was also replaced by parallelism on a high-performance…
